Core components

The low-level functions and classes described here are part of the hiPhive core components and not intended to be used directly during normal operation.


hiphive.core.cluster_space_builder.build_cluster_space(cluster_space, prototype_structure)[source]

The permutation list is an indexed fast lookup table for permutation vectors.




hiphive.core.get_clusters.create_neighbor_matrices(atoms, cutoff)[source]
hiphive.core.get_clusters.extend_cluster(cluster, order)[source]
hiphive.core.get_clusters.generate_geometrical_clusters(atoms, nPrim, cutoff, order)[source]
hiphive.core.get_clusters.get_clusters(atoms, cutoffs, nPrim, multiplicity=True, use_geometrical_order=False)[source]

Generate a list of all clusters in the atoms object which includes the center atoms with positions within the cell metric. The cutoff determines up to which order and range clusters should be generated.

With multiplicity set to True clusters like [0,0] and [3,3,4 etc will be generated. This is useful when doing force constants but not so much for cluster expansions.

The geometrical order is the total number of different atoms in the cluster. [0,0,1] would have geometrical order 2 and [1,2,3,4] would have order 4. If the key word is True the cutoff criteria will be based on the geometrical order of the cluster. This is based on the observation that many body interactions decrease fast with cutoff but anharmonic interactions can be quite long ranged.

  • atoms (ASE Atoms object) – can be a general atoms object but must have pbc=False.
  • cutoffs (dict) – the keys specify the order while the values specify the cutoff radii
  • multiplicity (bool) – includes clusters where same atom appears more than once
  • geometrical_order (bool) – specifies if the geometrical order should be used as cutoff_order, otherwise the normal order of the cluster is used

a list of clusters where each entry is a tuple of indices, which refer to the atoms in the input supercell

Return type:

list of tuples


hiphive.core.get_orbits.generate_translated_clusters(cluster, extended_atoms)[source]
hiphive.core.get_orbits.get_orbits(cluster_list, atom_list, rotation_matrices, translation_vectors, permutations, prim)[source]

Generate a list of the orbits for the clusters in a supercell configuration.

This method requires as input a list of the clusters in a supercell configuration as well as a set of symmetry operations (rotations and translations). From this information it will generate a list of the orbits, i.e. the set of symmetry inequivalent clusters each associated with its respective set of equivalent clusters.

  • cluster_list (BiMap object) – a list of clusters
  • atom_list (BiMap object) – a list of atoms in a supercell
  • rotation_matrices (list of NumPy (3,3) arrays) – rotational symmetries to be imposed (e.g., from spglib)
  • translation_vectors (list of NumPy (3) arrays) – translational symmetries to be imposed (e.g., from spglib)
  • permutations (list of permutations) – lookup table for permutations
  • prim (hiPhive Atoms object) – primitive structure

orbits associated with the list of input clusters

Return type:

list of Orbits objs

hiphive.core.get_orbits.get_permutation_map(atoms, rotations, translations, basis)[source]
hiphive.core.get_orbits.populate_orbit(orbit, permutations, clusters, cluster, permutation_map, extended_atoms, cluster_is_found)[source]


Contains the Orbit class which hold onformation about equivalent clusters.

class hiphive.core.orbit.Orbit[source]

This class serves as a container for storing data pertaining to an orbit.


list of OrientationFamily objs – The orientation families of the orbit


list of tuples – Each eigensymmetry correspond to a pair where the first index is the symmetry and the second is the permutation


list of NumPy arrays – The decomposition of the force constant into symmetry elements


int – index of cluster that serves as prototype for this orbit

In the code the first symmetry is always the identity so the first orientation family should always correspond to the prototype

static read(f)[source]

Load a ClusterSpace from file

Parameters:f (string or file object) – name of input file (string) or stream to load from (file object)

Write a Orbit instance to a file.

Parameters:f (str or file object) – name of input file (str) or stream to write to (file object)

Compute the geometrical size of a 3-dimensional point cloud. The geometrical size is defined as the average distance to the geometric center.

Parameters:positions (list of 3-dimensional vectors) – positions of points in cloud
Returns:geometric size of point cloud
Return type:float

Compute the maximum distance between any two points in a 3-dimensional point cloud. This is equivalent to the “size” criterion used when imposing a certain (pair) cutoff criterion during construction of a set of clusters.

Parameters:positions (list of 3-dimensional vectors) – positions of points in cloud
Returns:maximum distance betwee any two points
Return type:float


The orientation_family module contains the OrientationFamily class, which acts like a struct that contains information about a set of equal clusters of equal orientation in space.

class hiphive.core.orientation_family.OrientationFamily(symmetry_index=None)[source]

A container for storing information for a “family of orientations”.

An orbit contains many clusters. Some of the clusters can be tranlsated onto each other and other must first be rotated. A set of clusters in the orbit which can all be translated onto each other are oriented in the same way and belongs to the same orientation family. The family is haracterized by the symmetry (rotation) which relates it to the prototype structure of the orbit.

Since the clusters are generally stored sorted the permutation information must also be stored.

Parameters:symmetry_index (int) – The index of the symmetry corresponding to spglibs symmetry

int – The index of the symmetry corresponding to spglibs symmetry


list of ints – The indices of the clusters belonging to this family


list of ints – The indices of the permutation vector

static read(f)[source]

Load a OrientationFamily object from a pickle file.

Parameters:f (str or file object) – name of input file (str) or stream to load from (file object)
Return type:OrientationFamily object

Write the object to file.

Parameters:f (str or file object) – name of input file (str) or stream to write to (file object)


hiphive.core.eigentensors.assemble_new_eigentensor(eigentensors, solution)[source]

Ensum can be optimized if the shapes and contraction order of the inputs are known beforehand.


This function creates eigentensor which fulfill the atom label symmetry. The only information needed is the corresponding cluster

hiphive.core.eigentensors.populate_constraint_matrix(M, ets, R, p_inv, einsum_path)[source]


Module containing tensor related functions

hiphive.core.tensors.get_einsum_path(indices_list, optimize='optimal')[source]
hiphive.core.tensors.rotate_tensor(T, R, path=None)[source]

Equivalent to T_abc... = T_ijk... R_ia R_jb R_kc ...

hiphive.core.tensors.rotation_to_cart_coord(R, cell)[source]

Return the rotation matrix in cart coord given a cell metric


Collection of functions and classes for handling information concerning atoms and structures, including the relationship between primitive cell and supercells that are derived thereof.

class hiphive.core.atoms.Atom(*args)[source]

Unique representation of an atom in a lattice with a basis

Class for storing information about the position of an atom in a supercell relative to the origin of the underlying primitive cell. This class is used for handling the relationship between a primitive cell and supercells derived thereof. Initialize self


An Atom object can be initialized by providing either

  • four elements, which are interpreted as site, offset_x, offset_y, offset_z
  • a list with four elements, which is interpreted as [site, offset_x, offset_y, offset_z], or
  • a list with two elements, which is interpreted as [site, [offset_x, offset_y, offset_z]].

where all of the elements are integers. Here, site indicates the index of a site in the primitive basis and offset describes the translation of the atom position in the supercell relative to the origin in units of the primitive cell vectors.


The three following commands yield identical results

>>> atom = Atom(1, 0, 0, 3)
>>> atom = Atom(1, [0, 0, 3])
>>> atom = Atom([1, 0, 0, 3])

list of ints – translational offset of the supercell site relative to the origin of the primitive cell in units of primitive lattice vectors

pos(atom, basis, cell)[source]

int – index of corresponding site in the primitive basis

spos(atom, basis)[source]
static spos_to_atom(spos, basis, tol=None)[source]
class hiphive.core.atoms.Atoms(symbols=None, positions=None, numbers=None, tags=None, momenta=None, masses=None, magmoms=None, charges=None, scaled_positions=None, cell=None, pbc=None, celldisp=None, constraint=None, calculator=None, info=None)[source]

Minimally augmented version of the ASE Atoms class suitable for handling primitive cell information.

Saves and loads by pickle.


NumPy array – scaled coordinates of the sites in the primitive basis

static read(f)[source]

Load an hiPhive Atoms object from file.

Parameters:f (str or file object) – name of input file (str) or stream to load from (file object)
Return type:hiPhive Atoms object

Write the object to file.

Note: Only the cell, basis and numbers are stored!

Parameters:f (str or file object) – name of input file (str) or stream to write to (file object)
hiphive.core.atoms.align_super_cell(sc, prim, symprec=None)[source]

Rotate and translate a supercell configuration such that it is aligned with the target primitive cell.

  • sc (ASE Atoms object) – supercell configuration
  • prim (ASE Atoms object) – target primitive configuration
  • symprec (float) – precision parameter forwarded to spglib

The method returns the aligned supercell configuration as well as the rotation matrix and the translation vector that related the input to the aligned supercell configuration.

Return type:

tuple of ASE Atoms object, NumPy (3,3) array, 3-dimensional vector

hiphive.core.atoms.atom_to_spos(atom, basis)[source]

Helper function for obtaining the position of a supercell atom in scaled coordinates.

  • atom (hiPhive Atom object) – supercell atom
  • basis (list) – positions of sites in the primitive basis

array – scaled coordinates of an atom in a supercell

Return type:

NumPy array

hiphive.core.atoms.spos_to_atom(spos, basis, tol=0.0001)[source]

Helper function for transforming a supercell position to the primitive basis.

  • spos (list) – scaled coordinates of an atom in a supercell
  • basis (list) – positions of sites in the primitive basis
  • tol (float) – a general tolerance

supercell atom

Return type:

hiPhive Atom object


hiphive.core.relate_structures.atoms_equal(atoms_1, atoms_2)[source]

Compares two configurations with periodic boundary conditions.

The atoms should have the same * cell * elements * scaled positions (mod 1) * pbc

Compared to Atoms.__eq__ the order of the atoms does not matter

hiphive.core.relate_structures.get_primitive_cell(atoms, to_primitive=True, no_idealize=True)[source]

Get primitive cell from spglib

  • atoms (ASE Atoms object) –
  • to_primitive (bool) – passed to spglib
  • no_idealize (bool) – passed to spglib
hiphive.core.relate_structures.is_rotation(R, cell_metric=None)[source]

Checks if rotation matrix is orthonormal

A cell metric can be passed of the rotation matrix is in scaled coordinates

  • R (Numpy 3x3 matrix) – The rotation matrix
  • cell_metric (Numpy 3x3 matrix) – Optinal cell metric if the rotation is in scaled coordinates
hiphive.core.relate_structures.relate_structures(reference, target)[source]

Finds rotation and translation operations that align two structures with periodic boundary conditions.

The rotation and translation in Cartesian coordinates will map the reference structure onto the target

  • reference (ASE Atoms) – The reference structure to be mapped
  • target (ASE Atoms) – The target structure

  • R (NumPy (3, 3) array) – The rotation matrix in Cartesian coordinates
  • T (NumPy (3) array) – The translation vector in Cartesian coordinates

hiphive.core.relate_structures.rotate_atoms(atoms, rotation)[source]

Rotates the cell and positions of Atoms and returns a copy

  • atoms (ASE Atoms object) –
  • rotation (Numpy 3x3 matrix) –


The utilities module contains various support functions and classes.

class hiphive.core.utilities.Progress(tot=None, mode='frac', estimate_remaining=True)[source]

Progress bar like functionality.



hiphive.core.core_config.constraint_vectors_compress_mode = 'symbolic'

True, False

hiphive.core.core_config.constraint_vectors_simplify_before_compress = True

‘symbolic’, ‘numeric’, None

hiphive.core.core_config.constraint_vectors_simplify_before_solve = True

‘symbolic’, ‘numeric’

class hiphive.core.core_config.core_config[source]
class eigensymmetries[source]
crystal_symmetries = True
class iterative[source]
method = 'symbolic'
simplify = True
simplify_tolerance = None
zero_tolerance = 1e-12
method = 'iterative'
rotation_integer_tolerance = 1e-12
hiphive.core.core_config.eigentensor_compress_before_solve = None

This might make the nullspace() more stable True, False

hiphive.core.core_config.eigentensor_compress_mode = None

If this is True, before every symbolic compression the values will be simplified by sympy, potentially turning them into exact rational or irrational numbers. This can be useful for systems with non-integer rotation matrices in cartesian coordinates e.g. hcp. The main purpose is to make the rref more stable against repeating rounding errors. True, False

hiphive.core.core_config.eigentensor_simplify_before_compress = False

If non compress was used during construction but used before solving True, False

hiphive.core.core_config.eigentensor_simplify_before_last_compress = False

If the compress_mode is None the constraint matrix might be compressed right before the nullspace() solver ‘numeric’, ‘symbolic’, None

hiphive.core.core_config.eigentensor_simplify_before_solve = True

‘symbolic’, ‘numeric’

hiphive.core.core_config.eigentensor_solve_mode = 'symbolic'

‘symbolic’, ‘numeric’

hiphive.core.core_config.integer_tolerance = 1e-12

For each symmetry, the constraint matrix can be reduced to square again. This can be done either by ‘symbolic’, ‘numeric’ or not at all (None). Default is None since the matrix is often small enough to fit in memory. ‘symbolic’, ‘numeric’, None

hiphive.core.core_config.sum_rule_constraint_mode = 'symbolic'

True, False

hiphive.core.core_config.sum_rule_constraint_simplify = True

True, False