ovito.data

This Python module defines various data object types, which are produced and processed within OVITO’s data pipeline system. It also provides the DataCollection class as a container for such data objects as well as several utility classes for computing neighbor lists and iterating over the bonds of connected to a particle.

Data containers:

Data objects:

  • Property - uniform array of property values

  • SimulationCell - simulation box geometry and boundary conditions

  • SurfaceMesh - polyhedral mesh representing the boundaries of spatial regions

  • TriangleMesh - general mesh structure made of vertices and triangular faces

  • DislocationNetwork - set of discrete dislocation lines with Burgers vector information

Auxiliary data objects:

  • ElementType - base class for type descriptors used in typed properties

  • ParticleType - describes a single particle or atom type

  • BondType - describes a single bond type

Utility classes:

class ovito.data.BondType

Base: ovito.data.ElementType

Represents a bond type. This class inherits all its fields from the ElementType base class.

You can enumerate the list of defined bond types by accessing the bond_types bond property object:

bond_type_property = data.particles.bonds.bond_types
for t in bond_type_property.types:
    print(t.id, t.name, t.color, t.radius)
property radius

This attribute controls the display radius of all bonds of this type.

When set to zero, bonds of this type will be rendered using the standard width specified by the BondsVis.radius parameter. Furthermore, precedence is given to any per-bond widths assigned to the Width bond property if that property exists.

Default:

0.0

New in version 3.10.0.

class ovito.data.Bonds

Base: ovito.data.PropertyContainer

Stores the list of bonds and their properties. A Bonds object is always part of a parent Particles object. You can access it as follows:

data = pipeline.compute()
print("Number of bonds:", data.particles.bonds.count)

The Bonds class inherits the count attribute from its PropertyContainer base class. This attribute returns the number of bonds.

Bond properties

Bonds can be associated with arbitrary bond properties, which are managed in the Bonds container as a set of Property data arrays. Each bond property has a unique name by which it can be looked up:

print("Bond property names:")
print(data.particles.bonds.keys())
if 'Length' in data.particles.bonds:
    length_prop = data.particles.bonds['Length']
    assert(len(length_prop) == data.particles.bonds.count)

New bond properties can be added using the PropertyContainer.create_property() method.

Bond topology

The Topology bond property, which is always present, defines the connectivity between particles in the form of a N x 2 array of indices into the Particles array. In other words, each bond is defined by a pair of particle indices.

for a,b in data.particles.bonds.topology:
    print("Bond from particle %i to particle %i" % (a,b))

Note that the bonds of a system are not stored in any particular order. If you need to enumerate all bonds connected to a certain particle, you can use the BondsEnumerator utility class for that.

Bonds visualization

The Bonds data object has a BondsVis element attached to it, which controls the visual appearance of the bonds in rendered images. It can be accessed through the vis attribute:

data.particles.bonds.vis.enabled = True
data.particles.bonds.vis.flat_shading = True
data.particles.bonds.vis.width = 0.3

Computing bond vectors

Since each bond is defined by two indices into the particles array, we can use these indices to determine the corresponding spatial bond vectors connecting the particles. They can be computed from the positions of the particles:

topology = data.particles.bonds.topology
positions = data.particles.positions
bond_vectors = positions[topology[:,1]] - positions[topology[:,0]]

Here, the first and the second column of the bonds topology array are used to index into the particle positions array. The subtraction of the two indexed arrays yields the list of bond vectors. Each vector in this list points from the first particle to the second particle of the corresponding bond.

Finally, we may have to correct for the effect of periodic boundary conditions when a bond connects two particles on opposite sides of the box. OVITO keeps track of such cases by means of the the special Periodic Image bond property. It stores a shift vector for each bond, specifying the directions in which the bond crosses periodic boundaries. We make use of this information to correct the bond vectors computed above. This is done by adding the product of the cell matrix and the shift vectors from the Periodic Image bond property:

bond_vectors += numpy.dot(data.cell[:3,:3], data.particles.bonds.pbc_vectors.T).T

The shift vectors array is transposed here to facilitate the transformation of the entire array of vectors with a single 3x3 cell matrix. To summarize: In the two code snippets above, we have performed the following calculation of the unwrapped vector \(\mathbf{v}\) for every bond (a, b) in parallel:

\(\mathbf{v} = \mathbf{x}_b - \mathbf{x}_a + \mathbf{H} \cdot (n_x, n_y, n_z)^{T}\),

with \(\mathbf{H}\) denoting the simulation cell matrix and \((n_x, n_y, n_z)\) the bond’s PBC shift vector.

Standard bond properties

The following standard properties are defined for bonds:

Property name

Python access

Data type

Component names

Bond Type

bond_types

int32

Color

colors

float32

R, G, B

Length

float64

Particle Identifiers

int64

A, B

Periodic Image

pbc_vectors

int32

X, Y, Z

Selection

selection

int8

Topology

topology

int64

1, 2

Transparency

float32

Width

float32

add_bond(a, b, type=None, pbcvec=None)

Creates a new bond between two particles a and b, both parameters being indices into the particles list.

Parameters:
  • a (int) – Index of first particle connected by the new bond. Particle indices start at 0.

  • b (int) – Index of second particle connected by the new bond.

  • type (int) – Optional type ID to be assigned to the new bond. This value will be stored to the bond_types array.

  • pbcvec (tuple) – Three integers specifying the bond’s crossings of periodic cell boundaries. The information will be stored in the pbc_vectors array.

Returns:

The index of the newly created bond, i.e. (Bonds.count-1).

The method does not check if there already is an existing bond connecting the same pair of particles.

The method does not check if the particle indices a and b do exist. Thus, it is your responsibility to ensure that both indices are in the range 0 to (Particles.count-1).

In case the SimulationCell has periodic boundary conditions enabled, and the two particles connected by the bond are located in different periodic images, make sure you provide the pbcvec argument. It is required so that OVITO does not draw the bond as a direct line from particle a to particle b but as a line passing through the periodic cell faces. You can use the Particles.delta_vector() function to compute pbcvec or use the pbc_shift vector returned by the CutoffNeighborFinder utility.

property bond_types

The Property data array for the Bond Type standard bond property; or None if that property is undefined.

property colors

The Property data array for the Color standard bond property; or None if that property is undefined.

property pbc_vectors

The Property data array for the Periodic Image standard bond property; or None if that property is undefined.

property selection

The Property data array for the Selection standard bond property; or None if that property is undefined.

property topology

The Property data array for the Topology standard bond property; or None if that property is undefined.

class ovito.data.BondsEnumerator(bonds: ovito.data.Bonds)

Utility class that permits efficient iteration over the bonds connected to specific particles.

The constructor takes a Bonds object as input. From the generally unordered list of bonds, the BondsEnumerator will build a lookup table for quick enumeration of bonds of particular particles.

All bonds connected to a specific particle can be subsequently visited using the bonds_of_particle() method.

Warning: Do not modify the underlying Bonds object while the BondsEnumerator is in use. Adding or deleting bonds would render the internal lookup table of the BondsEnumerator invalid.

Usage example

from ovito.io import import_file
from ovito.data import BondsEnumerator
from ovito.modifiers import ComputePropertyModifier

# Load a dataset containing atoms and bonds.
pipeline = import_file('input/bonds.data.gz', atom_style='bond')

# For demonstration purposes, let's define a compute modifier that calculates the length 
# of each bond, storing the results in a new bond property named 'Length'.
pipeline.modifiers.append(ComputePropertyModifier(operate_on='bonds', output_property='Length', expressions=['BondLength']))

# Obtain pipeline results.
data = pipeline.compute()
positions = data.particles.positions  # array with atomic positions
bond_topology = data.particles.bonds.topology  # array with bond topology
bond_lengths = data.particles.bonds['Length']     # array with bond lengths

# Create bonds enumerator object.
bonds_enum = BondsEnumerator(data.particles.bonds)

# Loop over atoms.
for particle_index in range(data.particles.count):
    # Loop over bonds of current atom.
    for bond_index in bonds_enum.bonds_of_particle(particle_index):
        # Obtain the indices of the two particles connected by the bond:
        a = bond_topology[bond_index, 0]
        b = bond_topology[bond_index, 1]
        
        # Bond directions can be arbitrary (a->b or b->a):
        assert(a == particle_index or b == particle_index)
        
        # Obtain the length of the bond from the 'Length' bond property:
        length = bond_lengths[bond_index]

        print("Bond from atom %i to atom %i has length %f" % (a, b, length))
bonds_of_particle(index: int) Iterator[int]

Returns an iterator yielding the indices of the bonds connected to the given particle. The indices can be used to index into the Property arrays of the Bonds object.

class ovito.data.CutoffNeighborFinder(cutoff, data_collection)

A utility class that computes particle neighbor lists.

This class lets you iterate over all neighbors of a particle that are located within a specified spherical cutoff. You can use it to build neighbor lists or perform computations that require neighbor vector information.

The constructor takes a positive cutoff radius and a DataCollection providing the input particles and the SimulationCell (needed for periodic systems).

Once the CutoffNeighborFinder has been constructed, you can call its find() method to iterate over the neighbors of a particle, for example:

from ovito.io import import_file
from ovito.data import CutoffNeighborFinder

# Load input simulation file.
pipeline = import_file("input/simulation.dump")
data = pipeline.compute()

# Initialize neighbor finder object:
cutoff = 3.5
finder = CutoffNeighborFinder(cutoff, data)

# Prefetch the property array containing the particle type information:
ptypes = data.particles.particle_types

# Loop over all particles:
for index in range(data.particles.count):
    print("Neighbors of particle %i:" % index)

    # Iterate over the neighbors of the current particle:
    for neigh in finder.find(index):
        print(neigh.index, neigh.distance, neigh.delta, neigh.pbc_shift)

        # The index can be used to access properties of the current neighbor, e.g.
        type_of_neighbor = ptypes[neigh.index]

Note: In case you rather want to determine the N nearest neighbors of a particle, use the NearestNeighborFinder class instead.

Parameters:
find(index)

Returns an iterator over all neighbors of the given particle.

Parameters:

index (int) – The zero-based index of the central particle whose neighbors should be enumerated.

Returns:

A Python iterator that visits all neighbors of the central particle within the cutoff distance. For each neighbor the iterator returns an object with the following property fields:

  • index: The zero-based global index of the current neighbor particle.

  • distance: The distance of the current neighbor from the central particle.

  • distance_squared: The squared neighbor distance.

  • delta: The three-dimensional vector connecting the central particle with the current neighbor (taking into account periodicity).

  • pbc_shift: The periodic shift vector, which specifies how often each periodic boundary of the simulation cell is crossed when going from the central particle to the current neighbor.

The index value returned by the iterator can be used to look up properties of the neighbor particle, as demonstrated in the example above.

Note that all periodic images of particles within the cutoff radius are visited. Thus, the same particle index may appear multiple times in the neighbor list of the central particle. In fact, the central particle may be among its own neighbors in a small periodic simulation cell. However, the computed vector (delta) and PBC shift (pbc_shift) will be unique for each visited image of the neighbor particle.

find_all(indices=None, sort_by=None)

This is a vectorized version of the find() method, computing the neighbor lists and neighbor vectors of several particles in a single operation. Thus, this method can help you avoid a slow, nested Python loop in your code and it will make use of all available processor cores. You can request the neighbor lists for the whole system in one go, or just for a specific subset of particles given by indices.

The method produces a uniform array of neighbor list entries. Each entry comprises a pair of indices, i.e. the central particle and one of its neighboring particles within the cutoff distance, and the corresponding spatial neighbor vector in 3d Cartesian coordinates. For best performance, the method returns all neighbors of all particles as one large array, which is unsorted by default (sort_by = None). That means the neighbors of central particles will not form contiguous blocks in the output array; entries belonging to different central particles may rather appear in intermingled order!

Set sort_by to 'index' to request grouping the entries in the output array based on the central particle index. That means each particle’s neighbor list will be output as a contiguous block. All blocks are stored back-to-back in the output array in ascending order of the central particle index or, if parameter indices was specified, in that order. The ordering of neighbor entries within each block will still be arbitrary though. To change this, set sort_by to 'distance', which additionally sorts the neighbors of each particle by increasing distance.

The method returns two NumPy arrays:

neigh_idx : Array of shape (M, 2) containing pairs of indices of neighboring particles, with M equal to the total number of neighbors in the system. Note that the array will contain symmetric entries (a, b) and (b, a) if neighbor list computation was requested for both particles a and b and they are within reach of each other.

neigh_vec : Array of shape (M, 3) containing the xyz components of the Cartesian neighbor vectors (“delta”), which connect the M particle pairs stored in neigh_idx.

Parameters:
  • indices – List of zero-based indices of central particles for which the neighbor lists should be computed. If left unspecified, neighbor lists will be computed for every particle in the system.

  • sort_by – One of “index” or “distance”. Requests ordering of the output arrays based on central particle index and, optionally, neighbor distance. If left unspecified, neighbor list entries will be returned in completely arbitrary order.

Returns:

(neigh_idx, neigh_vec)

Tip

Sorting of neighbor lists will incur an additional runtime cost and should only be requested if necessary. In any case, however, this vectorized method will be much faster than an equivalent Python for-loop invoking the find() method for each individual particle.

Attention

The same index pair (a, b) may appear multiple times in the list neigh_idx if the SimulationCell uses periodic boundary conditions and its size is smaller than twice the neighbor cutoff radius. Note that, in such a case, the corresponding neighbor vectors in neigh_vec will still be unique, because they are computed for each periodic image of the neighbor b.

New in version 3.8.1.

find_at(coords)

Returns an iterator over all particles located within the spherical range of the given center position. In contrast to find() this method can search for neighbors around arbitrary spatial locations, which don’t have to coincide with any physical particle position.

Parameters:

coords – A (x,y,z) coordinate triplet specifying the center location around which to search for particles.

Returns:

A Python iterator enumerating all particles within the cutoff distance. For each neighbor the iterator returns an object with the following properties:

  • index: The zero-based global index of the current neighbor particle.

  • distance: The distance of the current particle from the center position.

  • distance_squared: The squared distance.

  • delta: The three-dimensional vector from the center to the current neighbor (taking into account periodicity).

  • pbc_shift: The periodic shift vector, which specifies how often each periodic boundary of the simulation cell is crossed when going from the center point to the current neighbor.

The index value returned by the iterator can be used to look up properties of the neighbor particle, as demonstrated in the example above.

Note that all periodic images of particles within the cutoff radius are visited. Thus, the same particle index may appear multiple times in the neighbor list. However, the computed vector (delta) and image offset (pbc_shift) will be unique for each visited image of a neighbor particle.

neighbor_distances(index)

Returns the list of distances between some central particle and all its neighbors within the cutoff range.

Parameters:

index (int) – The 0-based index of the central particle whose neighbors should be enumerated.

Returns:

NumPy array containing the radial distances to all neighbor particles within the cutoff range (in arbitrary order).

This method is equivalent to the following code, but performance is typically a lot better:

def neighbor_distances(index):
    distances = []
    for neigh in finder.find(index):
        distances.append(neigh.distance)
    return numpy.asarray(distances)
neighbor_vectors(index)

Returns the list of vectors from some central particle to all its neighbors within the cutoff range.

Parameters:

index (int) – The 0-based index of the central particle whose neighbors should be enumerated.

Returns:

Two-dimensional NumPy array containing the vectors to all neighbor particles within the cutoff range (in arbitrary order).

The method is equivalent to the following code, but performance is typically a lot better:

def neighbor_vectors(index):
    vecs = []
    for neigh in finder.find(index):
        vecs.append(neigh.delta)
    return numpy.asarray(vecs)
class ovito.data.DataCollection

Base: ovito.data.DataObject

A DataCollection is a container class holding together individual data objects, each representing different fragments of a dataset. For example, a dataset loaded from a simulation data file may consist of particles, the simulation cell information and additional auxiliary data such as the current timestep number of the snapshots, etc. All this information is contained in one DataCollection, which exposes the individual pieces of information as sub-objects, for example, via the DataCollection.particles, DataCollection.cell and DataCollection.attributes fields.

Data collections are the elementary entities that get processed within a data Pipeline. Each modifier receives a data collection from the preceding modifier, alters it in some way, and passes it on to the next modifier. The output data collection of the last modifier in the pipeline is returned by the Pipeline.compute() method.

A data collection essentially consists of a bunch of DataObjects, which are all stored in the DataCollection.objects list. Typically, you don’t access the data objects through this list directly but rather use one of the special accessor fields provided by the DataCollection class, which give more convenient access to data objects of a particular kind. For example, the surfaces dictionary provides key-based access to all the SurfaceMesh instances currently in the data collection.

You can programmatically add or remove data objects from a data collection by manipulating its objects list. For instance, to populate a new data collection instance that is initially empty with a new SimulationCell object:

data = DataCollection()
cell = SimulationCell()
data.objects.append(cell)
assert(data.cell is cell)
apply(modifier, frame=None)

This method applies a Modifier function to the data stored in this collection to modify it in place.

Parameters:
  • modifier (ovito.pipeline.Modifier) – The modifier object that should alter the contents of this data collection in place.

  • frame (int) – Optional animation frame number to be passed to the modifier function, which may use it for time-dependent modifications.

The method allows modifying a data collection with one of OVITO’s modifiers directly without the need to build up a complete Pipeline first. In contrast to a data pipeline, the apply() method executes the modifier function immediately and alters the data in place. In other words, the original data in this DataCollection gets replaced by the output produced by the invoked modifier function. It is possible to first create a copy of the original data using the clone() method if needed. The following code example demonstrates how to use apply() to successively modify a dataset:

from ovito.io import import_file
from ovito.modifiers import *

data = import_file("input/simulation.dump").compute()
data.apply(CoordinationAnalysisModifier(cutoff=2.9))
data.apply(ExpressionSelectionModifier(expression="Coordination<9"))
data.apply(DeleteSelectedModifier())

Note that it is typically possible to achieve the same result by first populating a Pipeline with the modifiers and then calling its compute() method at the very end:

pipeline = import_file("input/simulation.dump")
pipeline.modifiers.append(CoordinationAnalysisModifier(cutoff=2.9))
pipeline.modifiers.append(ExpressionSelectionModifier(expression="Coordination<9"))
pipeline.modifiers.append(DeleteSelectedModifier())
data = pipeline.compute()

An important use case of the apply() method is in the implementation of a user-defined modifier function, making it possible to invoke other modifiers as sub-routines:

# A user-defined modifier function that calls the built-in ColorCodingModifier
# as a sub-routine to assign a color to each atom based on some property
# created within the function itself:
def modify(frame: int, data: DataCollection):
    data.particles_.create_property('idx', data=numpy.arange(data.particles.count))
    data.apply(ColorCodingModifier(property='idx'), frame)

# Set up a data pipeline that uses the user-defined modifier function:
pipeline = import_file("input/simulation.dump")
pipeline.modifiers.append(modify)
data = pipeline.compute()
property attributes

This field contains a dictionary view with all the global attributes currently associated with this data collection. Global attributes are key-value pairs that represent small tokens of information, typically simple value types such as int, float or str. Every attribute has a unique identifier such as 'Timestep' or 'ConstructSurfaceMesh.surface_area'. This identifier serves as lookup key in the attributes dictionary. Attributes are dynamically generated by modifiers in a data pipeline or come from the data source. For example, if the input simulation file contains timestep information, the timestep number is made available by the FileSource as the 'Timestep' attribute. It can be retrieved from pipeline’s output data collection:

>>> pipeline = import_file('snapshot_140000.dump')
>>> pipeline.compute().attributes['Timestep']
140000

Some modifiers report their calculation results by adding new attributes to the data collection. See each modifier’s reference documentation for the list of attributes it generates. For example, the number of clusters identified by the ClusterAnalysisModifier is available in the pipeline output as an attribute named ClusterAnalysis.cluster_count:

pipeline.modifiers.append(ClusterAnalysisModifier(cutoff = 3.1))
data = pipeline.compute()
nclusters = data.attributes["ClusterAnalysis.cluster_count"]

The ovito.io.export_file() function can be used to output dynamically computed attributes to a text file, possibly as functions of time:

export_file(pipeline, "data.txt", "txt/attr",
    columns = ["Timestep", "ClusterAnalysis.cluster_count"],
    multiple_frames = True)

If you are writing your own modifier function, you let it add new attributes to a data collection. In the following example, the CommonNeighborAnalysisModifier first inserted into the pipeline generates the 'CommonNeighborAnalysis.counts.FCC' attribute to report the number of atoms that have an FCC-like coordination. To compute an atomic fraction from that, we need to divide the count by the total number of atoms in the system. To this end, we append a user-defined modifier function to the pipeline, which computes the fraction and outputs the value as a new attribute named 'fcc_fraction'.

pipeline.modifiers.append(CommonNeighborAnalysisModifier())
            
def compute_fcc_fraction(frame, data):
    n_fcc = data.attributes['CommonNeighborAnalysis.counts.FCC']
    data.attributes['fcc_fraction'] = n_fcc / data.particles.count

pipeline.modifiers.append(compute_fcc_fraction)
print(pipeline.compute().attributes['fcc_fraction'])
property cell

Returns the SimulationCell data object describing the cell vectors and periodic boundary condition flags. It may be None.

Important

The SimulationCell data object returned by this attribute may be marked as read-only, which means your attempts to modify the cell object will raise a Python error. This is typically the case if the data collection was produced by a pipeline and its objects are owned by the system.

If you intend to modify the SimulationCell data object within this data collection, use the cell_ attribute instead to explicitly request a mutable version of the cell object. See topic Announcing object modification for more information. Use cell for read access and cell_ for write access, e.g.

print(data.cell.volume)
data.cell_.pbc = (True, True, False)

To create a SimulationCell in a data collection that might not have a simulation cell yet, use the create_cell() method or simply assign a new instance of the SimulationCell class to the cell attribute.

clone()

Returns a copy of this DataCollection containing the same data objects as the original.

The method may be used to retain a copy of the original data before modifying a data collection in place, for example using the apply() method:

original = data.clone()
data.apply(ExpressionSelectionModifier(expression="Position.Z < 0"))
data.apply(DeleteSelectedModifier())
print("Number of atoms before:", original.particles.count)
print("Number of atoms after:", data.particles.count)

Note that the clone() method performs an inexpensive, shallow copy, meaning that the newly created collection will still share the data objects with the original collection. Data objects that are shared by two or more data collections are protected against modification by default to avoid unwanted side effects. Thus, in order to subsequently modify the data objects in either the original collection or its copy, you will have to use the underscore notation or the DataObject.make_mutable() method to explicitly request a deep copy of the particular data object(s) you want to modify. For example:

copy = data.clone()
# Data objects are shared by original and copy:
assert(copy.cell is data.cell)

# In order to modify the SimulationCell in the dataset copy, we must request
# a mutable version of the SimulationCell using the 'cell_' accessor:
copy.cell_.pbc = (False, False, False)

# As a result, the cell object in the second data collection has been replaced
# with a deep copy and the two data collections no longer share the same
# simulation cell object:
assert(copy.cell is not data.cell)
create_cell(matrix, pbc=(True, True, True), vis_params=None)

This convenience method conditionally creates a new SimulationCell object and stores it in this data collection. If a simulation cell already existed in the collection (cell is not None), then that cell object is replaced with a modifiable copy if necessary and the matrix and PBC flags are set to the given values. The attached SimulationCellVis element is maintained in this case.

Parameters:
  • matrix – A 3x4 array to initialize the cell matrix with. It specifies the three cell vectors and the origin.

  • pbc – A tuple of three Booleans specifying the cell’s pbc flags.

  • vis_params (Mapping[str, Any]) – Optional dictionary to initialize attributes of the attached SimulationCellVis element (only used if the cell object is newly created by the method).

Return type:

ovito.data.SimulationCell

The logic of this method is roughly equivalent to the following code:

def create_cell(data: DataCollection, matrix, pbc, vis_params=None) -> SimulationCell:
    if data.cell is None:
        data.cell = SimulationCell(pbc=pbc)
        data.cell[...] = matrix
        data.cell.vis.line_width = <...> # Some value that scales with the cell's size
        if vis_params:
            for name, value in vis_params.items(): setattr(data.cell.vis, name, value)
    else:
        data.cell_[...] = matrix
        data.cell_.pbc = pbc
    return data.cell_

New in version 3.7.4.

create_particles(vis_params=None, **params)

This convenience method conditionally creates a new Particles container object and stores it in this data collection. If the data collection already contains an existing particles object (particles is not None), then that particles object is replaced with a modifiable copy if necessary. The associated ParticlesVis element is preserved.

Parameters:
  • params – Key/value pairs passed to the method as keyword arguments are used to set attributes of the Particles object (even if the particles object already existed).

  • vis_params (Mapping[str, Any]) – Optional dictionary to initialize attributes of the attached ParticlesVis element (only used if the particles object is newly created by the method).

Return type:

ovito.data.Particles

The logic of this method is roughly equivalent to the following code:

def create_particles(data: DataCollection, vis_params=None, **params) -> Particles:
    if data.particles is None:
        data.particles = Particles()
        if vis_params:
            for name, value in vis_params.items(): setattr(data.particles.vis, name, value)
    for name, value in params.items(): setattr(data.particles_, name, value)
    return data.particles_

Usage example:

coords = [(-0.06,  1.83,  0.81), # xyz coordinates of the 3 particle system to create
          ( 1.79, -0.88, -0.11),
          (-1.73, -0.77, -0.61)]
particles = data.create_particles(count=len(coords), vis_params={'radius': 1.4})
particles.create_property('Position', data=coords)

New in version 3.7.4.

property dislocations

Returns the DislocationNetwork data object; or None if there is no object of this type in the collection. Typically, the DislocationNetwork is created by a pipeline containing the DislocationAnalysisModifier.

property grids

Returns a dictionary view providing key-based access to all VoxelGrids in this data collection. Each VoxelGrid has a unique identifier key, which allows you to look it up in this dictionary. To find out which voxel grids exist in the data collection and what their identifiers are, use

print(data.grids)

Then retrieve the desired VoxelGrid from the collection using its identifier key, e.g.

charge_density_grid = data.grids['charge-density']
print(charge_density_grid.shape)

The view provides the convenience method grids.create(), which inserts a newly created VoxelGrid into the data collection. The method expects the unique identifier of the new grid as first argument. All other keyword arguments are forwarded to the constructor to initialize the member fields of the VoxelGrid class:

grid = data.grids.create(
    identifier="grid",
    title="Field",
    shape=(10,10,10),
    domain=data.cell)

If there is already an existing grid with the same identifier in the collection, the create() method modifies and returns that existing grid instead of creating another one.

property lines

A dictionary view providing key-based access to all Lines objects in this data collection. Each Lines object has a unique identifier key, which can be used to look it up in the dictionary. You can use

print(data.lines)

to see which identifiers exist. Then retrieve the desired Lines object from the collection using its identifier key, e.g.

lines = data.lines["trajectories"]
print(lines["Position"])

The Lines object with the identifier "trajectories", for example, is the one that gets created by the GenerateTrajectoryLinesModifier.

If you would like to create a new Lines object, in a user-defined modifier for instance, the dictionary view provides the method lines.create(), which creates a new Lines and adds it to the data collection. The method expects the unique identifier of the new lines object as first argument. All other keyword arguments are forwarded to the class constructor to initialize the member fields of the Lines object:

lines = data.lines.create(identifier="mylines")

If there is already an existing Lines object with the same identifier in the collection, the create() method returns that object instead of creating another one and makes sure it can be safely modified.

property objects

Flat list of all top-level DataObjects currently in this data collection. You can add or remove data objects from this list as needed.

Typically, however, you don’t need to work with this list directly, because the DataCollection class provides several convenience accessor attributes for the different flavors of data objects in OVITO. For example, DataCollection.particles returns the Particles object (by looking it up in the objects list for you). Dictionary-like views such as DataCollection.tables and DataCollection.surfaces provide key-based access to a particular class of data objects in the collection.

To add new objects to the data collection, you can append them to the objects list or, more conveniently, use creation functions such as create_particles(), create_cell(), or tables.create(), which are provided by the DataCollection class.

property particles

Returns the Particles object, which manages all per-particle properties. It may be None if the data collection contains no particle model at all.

Important

The Particles data object returned by this attribute may be marked as read-only, which means attempts to modify its contents will raise a Python error. This is typically the case if the data collection was produced by a pipeline and all data objects are owned by the system.

If you intend to modify the contents of the Particles object in some way, use the particles_ attribute instead to explicitly request a mutable version of the particles object. See topic Announcing object modification for more information. Use particles for read access and particles_ for write access, e.g.

print(data.particles.positions[0])
data.particles_.positions_[0] += (0.0, 0.0, 2.0)

To create a new Particles object in a data collection that might not have particles yet, use the create_particles() method or simply assign a new instance of the Particles class to the particles attribute.

property surfaces

Returns a dictionary view providing key-based access to all SurfaceMesh objects in this data collection. Each SurfaceMesh has a unique identifier key, which can be used to look it up in the dictionary. See the documentation of the modifier producing the surface mesh to find out what the right key is, or use

print(data.surfaces)

to see which identifier keys exist. Then retrieve the desired SurfaceMesh object from the collection using its identifier key, e.g.

surface = data.surfaces['surface']
print(surface.vertices['Position'])

The view provides the convenience method surfaces.create(), which inserts a newly created SurfaceMesh into the data collection. The method expects the unique identifier of the new surface mesh as first argument. All other keyword arguments are forwarded to the constructor to initialize the member fields of the SurfaceMesh class:

mesh = data.surfaces.create(
    identifier="surface",
    title="A surface mesh",
    domain=data.cell)

If there is already an existing mesh with the same identifier in the collection, the create() method modifies and returns that existing mesh instead of creating another one.

property tables

A dictionary view of all DataTable objects in this data collection. Each DataTable has a unique identifier key, which allows it to be looked up in this dictionary. Use

print(data.tables)

to find out which table identifiers are present in the data collection. Then use the identifier to retrieve the desired DataTable from the dictionary, e.g.

rdf = data.tables['coordination-rdf']
print(rdf.xy())

The view provides the convenience method tables.create(), which inserts a newly created DataTable into the data collection. The method expects the unique identifier of the new data table as first argument. All other keyword arguments are forwarded to the constructor to initialize the member fields of the DataTable class:

# Code example showing how to compute a histogram of the particles' x-coordinates within some interval.
x_interval = (0.0, 100.0)
x_coords = data.particles.positions[:,0]
histogram = numpy.histogram(x_coords, bins=50, range=x_interval)[0]

# Output the histogram as a new DataTable, which makes it appear in OVITO's data inspector panel:
table = data.tables.create(
    identifier='binning',
    title='Binned particle counts',
    plot_mode=DataTable.PlotMode.Histogram,
    interval=x_interval,
    axis_label_x='Position X',
    count=len(histogram))
table.y = table.create_property('Particle count', data=histogram)

If there is already an existing table with the same identifier in the collection, the create() method modifies and returns that existing table instead of creating another one.

property triangle_meshes

This is a dictionary view providing key-based access to all TriangleMesh objects currently stored in this data collection. Each TriangleMesh has a unique identifier key, which can be used to look it up in the dictionary.

class ovito.data.DataObject

Abstract base class for all data object types in OVITO.

A DataObject represents a fragment of data processed in or by a data pipeline. See the ovito.data module for a list of different concrete data object types in OVITO. Data objects are typically contained in a DataCollection, which represents a whole data set. Furthermore, data objects can be nested into a hierarchy. For example, the Bonds data object is part of the parent Particles data object.

Data objects by themselves are non-visual objects. Visualizing the information stored in a data object in images is the responsibility of so-called visual elements. A data object may be associated with a DataVis element by assigning it to the data object’s vis field. Each type of visual element exposes a set of parameters that allow you to configure the appearance of the data visualization in rendered images and animations.

property identifier

The unique identifier string of the data object. It serves as lookup key in object dictionaries, for example the DataCollection.tables collection, or as a target name in various places where a data object needs to be referenced by name, e.g. the TimeAveragingModifier.operate_on field.

Data objects generated by modifiers in a pipeline typically have an automatically assigned identifier, as documented in the description of the respective modifier. When writing your own modifier function, you are responsible for giving new data objects created by your modifier function a meaningful identifier, so that subsequent modifiers in the pipeline can refer to these data objects.

make_mutable(subobj)

This helper method requests a deep copy of subobj, which must be a child DataObject of this parent DataObject. A copy will only be made in case the sub-object is currently referenced by at least one more parent object. If, however, the sub-object is exclusively owned by this DataObject, no copy is made and the original sub-object is returned as is. The returned object is safe to modify without unexpected side effects, because any shared ownership is converted to an exclusive ownership by the method.

Please see the section Announcing object modification for a discussion of object ownership and typical use cases for this method.

Parameters:

subobj (DataObject) – A existing sub-object of this parent data object, for which exclusive ownership is requested.

Returns:

A copy of subobj if its ownership was previously shared with some other parent. Otherwise the original object is returned.

property vis

The DataVis element currently associated with this data object, which is responsible for visually rendering the stored data. If set to None, the data object remains non-visual and doesn’t appear in rendered images or the viewports. Furthermore, note that the same DataVis element may be assigned to multiple data objects in order to synchronize their visual appearance.

class ovito.data.DataTable

Base: ovito.data.PropertyContainer

This data object type in OVITO represents a series of data points and is primarily used for histogram plots and other 2d graphs. More generally, however, it can store tabulated data consisting of an arbitrary number of columns of numeric values.

When used for 2d plots, a data table consists of an array of y-values and, optionally, an array of corresponding x-values, one value pair for each data point. These arrays are regular Property objects managed by the data table (a sub-class of PropertyContainer).

If no x data array has been set, the x-coordinates of the data points are implicitly determined by the table’s interval, which specifies a range along the x-axis over which the data points are evenly distributed. This is used, for example, for histograms with equisized bins, which don’t require explicit x-coordinates.

Data tables generated by modifiers such as CoordinationAnalysisModifier and HistogramModifier are accessible via the DataCollection.tables dictionary. You can retrieve them based on their unique identifier:

>>> print(data.tables)  # Print list of available data tables
{'coordination-rdf': DataTable(), 'clusters': DataTable()}

>>> rdf = data.tables['coordination-rdf']  # Look up tabulated RDF produced by a CoordinationAnalysisModifier

Exporting the values in a data table to a simple text file is possible using the export_file() function (use file format txt/table). You can either export a single DataTable or, as in the following code example, write a series of text files to export all the tables generated by a Pipeline for a simulation trajectory in one go. The key parameter selects which table from the DataCollection.tables dict is to be exported based on its unique identifier:

export_file(pipeline, 'output/rdf.*.txt', 'txt/table', key='coordination-rdf', multiple_frames=True)

To programatically create a new data table in Python, you should use the data.tables.create() method, for example when implementing a custom modifier function that should output its results as a data plot. The following code examples demonstrate how to add a new DataTable to the data collection and fill it with values.

To create a simple x-y scatter point plot:

# Create a DataTable object and specify its plot type and a human-readable title:
table = data.tables.create(identifier='myplot', plot_mode=DataTable.PlotMode.Scatter, title='My Scatter Plot')
# Set the x- and y-coordinates of the data points:
table.x = table.create_property('X coordinates', data=numpy.linspace(0.0, 10.0, 50))
table.y = table.create_property('Y coordinates', data=numpy.cos(table.x))

Note how the create_property() method is being used here to create two Property objects storing the coordinates of the data points. These property objects are then set as x and y arrays of the DataTable. This is necessary because a data table is a general PropertyContainer, which can store an arbitrary number of data columns. We have to tell the table which of these properties should be used as x- and y-coordinates for plotting.

A multi-line plot is obtained by using a vectorial property for the y array of the DataTable:

table = data.tables.create(identifier='plot', plot_mode=DataTable.PlotMode.Line, title='Trig functions')
table.x = table.create_property('Parameter x', data=numpy.linspace(0.0, 14.0, 100))
# Use the x-coords to compute two y-coords per data point: y(x) = (cos(x), sin(x))
y1y2 = numpy.stack((numpy.cos(table.x), numpy.sin(table.x)), axis=1)
table.y = table.create_property('f(x)', data=y1y2, components=['cos(x)', 'sin(x)'])

To generate a bar chart, the table’s x property must be filled with numeric IDs 0,1,2,3,… denoting the individual bars. Each bar is then given a text label by adding an ElementType to the Property.types list using Property.add_type_id():

table = data.tables.create(identifier='chart', plot_mode=DataTable.PlotMode.BarChart, title='My Bar Chart')
table.x = table.create_property('Structure Type', data=[0, 1, 2, 3])
table.x.add_type_id(0, table, name='Other')
table.x.add_type_id(1, table, name='FCC')
table.x.add_type_id(2, table, name='HCP')
table.x.add_type_id(3, table, name='BCC')
table.y = table.create_property('Count', data=[65, 97, 10, 75])

For histogram plots, one can specify the complete range of values covered by the histogram by setting the table’s interval property. The bin counts must be stored in the table’s y property. The number of elements in the y property array, together with the interval, determine the number of histogram bins and their uniform widths:

table = data.tables.create(identifier='histogram', plot_mode=DataTable.PlotMode.Histogram, title='My Histogram')
table.y = table.create_property('Counts', data=[65, 97, 10, 75])
table.interval = (0.0, 2.0)   # Four histogram bins of width 0.5 each.
table.axis_label_x = 'Values' # Set the x-axis label of the plot.

If you are going to access or export the data table after it was inserted into the DataCollection, refer to it using its unique identifier given at construction time, as shown in the following example:

def modify(frame: int, data: DataCollection):
    table = data.tables.create(identifier='trig-func', title='My Plot', plot_mode=DataTable.PlotMode.Line)
    table.x = table.create_property('X coords', data=numpy.linspace(0.0, 10.0, 50))
    table.y = table.create_property('Y coords', data=numpy.cos(frame * table.x))

pipeline.modifiers.append(modify)
export_file(pipeline, 'output/data.*.txt', 'txt/table', key='trig-func', multiple_frames=True)
property axis_label_x

The text label of the x-axis. This string is only used for a data plot if the x property of the data table is None and the x-coordinates of the data points are implicitly defined by the table’s interval property. Otherwise the name of the x property is used as axis label.

Default:

''

property interval

A pair of float values specifying the x-axis interval covered by the data points in this table. This interval is only used by the table if the data points do not possess explicit x-coordinates (i.e. if the table’s x property is None). In the absence of explicit x-coordinates, the interval specifies the range of equispaced x-coordinates implicitly generated by the data table.

Implicit x-coordinates are typically used in data tables representing histograms, which consist of equally-sized bins covering a certain value range along the x-axis. The bin size is then given by the interval width divided by the number of data points (see PropertyContainer.count property). The implicit x-coordinates of data points are placed in the centers of the bins. You can call the table’s xy() method to let it explicitly calculate the x-coordinates from the value interval for every data point.

Default:

(0.0, 0.0)

property plot_mode

The type of graphical plot for rendering the data in this DataTable. Must be one of the following predefined constants:

  • DataTable.PlotMode.NoPlot

  • DataTable.PlotMode.Line

  • DataTable.PlotMode.Histogram

  • DataTable.PlotMode.BarChart

  • DataTable.PlotMode.Scatter

Default:

DataTable.PlotMode.Line

property x

The Property containing the x-coordinates of the data points (for the purpose of plotting). The data points may not have explicit x-coordinates, so this property may be None for a data table. In such a case, the x-coordinates of the data points are implicitly determined by the table’s interval.

Default:

None

xy()

This convenience method returns a two-dimensional NumPy array containing both the x- and the y-coordinates of the data points in this data table. If the table has no explicit x coordinate property set, this method automatically computes equispaced x-coordinates from the interval.

property y

The Property containing the y-coordinates of the data points (for the purpose of plotting). This may be a vector property having more than one component per data point, in which case this data table represents a family of data plots.

Default:

None

class ovito.data.DislocationNetwork

Base: ovito.data.DataObject

A network of dislocation lines extracted from a crystal model by the DislocationAnalysisModifier. The modifier stores the dislocation network in a pipeline’s output data collection, from where it can be retrieved via the DataCollection.dislocations field:

data = pipeline.compute()
network = data.dislocations

The visual appareances of the dislocation lines in rendered images and videos is controlled by the associated DislocationVis element. You can access it as vis attribute of the DataObject base class:

network.vis.line_width = 1.5
network.vis.coloring_mode = DislocationVis.ColoringMode.ByBurgersVector

The lines list gives you access to the list of individual dislocations, which are all represented by instances of the DislocationNetwork.Line class. Furthermore, you can use the find_nodes() method to obtain a list of nodes at which dislocation lines are connected. These connections are represented by DislocationNetwork.Connector objects.

Important

Keep in mind that the list of dislocations is not ordered. In particular, the order in which the DXA modifier discovers each dislocation line in the crystal will change arbitrarily from one simulation frame to the next. Generally, there is no safe way to track individual dislocation lines through time, because dislocations (unlike atoms) don’t possess a unique identity and are not conserved – they can nucleate, annihilate, or undergo other reactions in between trajectory frames.

Code example

Complete script example for loading a molecular dynamics simulation, performing the DXA on a single snapshot, printing the list of extracted dislocation lines, and exporting the dislocation network to disk:

from ovito.io import import_file, export_file
from ovito.modifiers import DislocationAnalysisModifier
from ovito.data import DislocationNetwork

import ovito
ovito.enable_logging()

pipeline = import_file("input/simulation.dump")

# Extract dislocation lines from a crystal with diamond structure:
modifier = DislocationAnalysisModifier()
modifier.input_crystal_structure = DislocationAnalysisModifier.Lattice.CubicDiamond
pipeline.modifiers.append(modifier)
data = pipeline.compute()

total_line_length = data.attributes['DislocationAnalysis.total_line_length']
cell_volume = data.attributes['DislocationAnalysis.cell_volume']
print("Dislocation density: %f" % (total_line_length / cell_volume))

# Print list of dislocation lines:
print("Found %i dislocation lines" % len(data.dislocations.lines))
for line in data.dislocations.lines:
    print("Dislocation %i: length=%f, Burgers vector=%s" % (line.id, line.length, line.true_burgers_vector))
    print(line.points)

# Export dislocation lines to a CA file:
export_file(pipeline, "output/dislocations.ca", "ca")

# Or export dislocations to a ParaView VTK file:
export_file(pipeline, "output/dislocations.vtk", "vtk/disloc")

File import and export

Dislocation networks can be exported as a set of polylines to the legacy VTK file format using the ovito.io.export_file() function (specify the “vtk/disloc” format). During export to this file format, which does not support periodic boundary conditions, lines that cross a periodic domain boundary get split (i.e., wrapped around) at the simulation box boundaries.

OVITO’s native format for storing dislocation networks on disk is the CA file format, a simple text-based format that supports periodic boundary conditions. This format can be written by the export_file() function (”ca” format) and read by the import_file() function. It stores the dislocation lines, their connectivity, as well as the “defect mesh” produced by the DislocationAnalysisModifier.

class Connector
../_images/dislocation_connectors.png

A dislocation Line has two end point connectors and is described by a sequence of spatial points.

A connector object represents one of the two end points of each Line in the network. In other words, every Line has exactly two unique Connector objects belonging to the line. This pair is accessible via the Line.connectors attribute. Since dislocations always have a direction (their line sense, with respect to which their Burgers vector is defined), one connecter is located at the “head” (forward) and one at the “tail” (backward) end of the directed line.

../_images/dislocation_node.png

Three dislocation arms meet at a dislocation node (junction). The node is formed by a circular linked-list of connectors.

A dislocation network node (junction) is formed by several connectors located at the same point in space, as illustrated in the figure to the right. This node structure consists of three interlinked connectors belonging to the three dislocation arms meeting in the node. Dislocation arms can be either inbound or outbound.

Network nodes may consist of one, two, three or more connectors:

  1. A single connector, only interlinked with itself, represents a dangling line end. They occur when a dislocation terminates in another extended crystal defect, such as a grain boundary or free surface.

  2. A 2-node, consisting of two interlinked connectors, is part of a dislocation loop or infinite periodic dislocation line. They occur when a dislocation line is closed on itself, i.e, its head and tail are connected.

  3. A node with three or more connectors represents a physical dislocation junction, where three or more arms with non-zero Burgers vector meet.

The connectors belonging to the same network node are interlinked with each other in the form of a circular linked list. The Connector.next_arm field leads to the next connector in the circular list. The last connector of the node points back to the first connector of the node. This way, all connectors can be visited by starting from any connector of the node and following the Connector.next_arm field until the starting connector is reached again. For a 1-node (a dangling line end), the Connector.next_arm field points to itself.

The Connector.arms method yields a list of all connectors belonging to the same local node as this connector (including the connector itself). The Connector.arm_count property counts the number of connectors in the local node. The Connector.line field points to the Line object that the connector belongs to.

The DislocationNetwork.find_nodes() method can be used to generate a list of Connector objects, one for each node in the network. It is useful if you want to iterate over all unique nodes in the network.

New in version 3.10.2.

property arm_count

The number of arms meeting at the node formed by this connector and others, including the connector itself.

arms()

Returns a list of Connector objects representing the arms connected to the node. Each connector object links to a different dislocation line incident to the network node.

property is_head

True if the connector is located at the head of its dislocation line, i.e., self is self.line.connectors[1]. Then the connected Line is inbound on the node.

property is_tail

True if the connector is located at the tail of its dislocation line, i.e., self is self.line.connectors[0]. Then the connected Line is outbound from the node.

property line

The dislocation Line this connector belongs to.

property next_arm

The Connector belonging to the next dislocation line incident to the node.

property position

The Cartesian coordinates of the connector in the global simulation coordinate system. This corresponds to the start or end point of the dislocation, i.e., either self.line.points[0] or self.line.points[-1].

Note

The positions of the connectors in the same network node are typically identical, but they will differ if their dislocation lines belong to different periodic images of the simulation cell. In this case, the positions of the connectors are shifted by a periodicity vector of the simulation domain.

class Line

Describes a single continuous dislocation line that is part of a DislocationNetwork.

A dislocation line is a curve in 3d space, approximated by a sequence of points connected by linear line segments. You can query its total curve length or compute some location on the line from a linear path coordinate t using the method point_along_line(). The line is terminated by two connectors at its two end points, which represent the connectivity of the dislocation network.

A dislocation line is embedded in some crystallite (a region with uniform lattice orientation), which is identified by the numerical cluster_id. All dislocation lines belonging to the same crystallite share the same lattice coordinate system in which their true Burgers vectors are expressed. A line’s true_burgers_vector is given in Bravais lattice units.

Each crystallite has a particular mean orientation within the global simulation coordinate system and a mean lattice parameter and elastic strain. Applying these mean crystal properties to the true_burgers_vector yields the line’s spatial_burgers_vector, which is expressed in the global coordinate system shared by all dislocations of the entire DislocationNetwork. The spatial_burgers_vector is given in simulation coordinate units (typically Angstroms).

The is_loop property flag indicates that the two end points of the dislocation line form a 2-junction. This property does not necessarily mean that the dislocation forms an actual circular loop. In simulations using periodic boundary conditions, a straight dislocation can also connect to itself through the periodic cell boundaries and form an infinite periodic line. This situation is indicated by the is_infinite_line property, which implies that the is_loop property is also true.

All fields of this class are read-only. To modify a dislocation line, you can use the DislocationNetwork.set_line() method.

Changed in version 3.10.2: Renamed this class from DislocationSegment to DislocationNetwork.Line.

property cluster_id

The numeric identifier of the crystal cluster containing this dislocation line. Crystal cluster is the technical term for a contiguous group of atoms forming a spatial region with uniform lattice orientation, i.e., a crystallite or grain.

The true_burgers_vector of the dislocation is expressed in the local coordinate system of the crystal cluster. The spatial_burgers_vector of the dislocation is computed by transforming the true Burgers vector with the mean elastic deformation gradient tensor of the crystal cluster.

property connectors
../_images/dislocation_connectors.png

A tuple of two Connector objects representing the two end points of the dislocation line. The first connector is located at the start of the line (its tail), the second connector at the end of the line (its head).

New in version 3.10.2.

property custom_color

The RGB color value to be used for visualizing this particular dislocation line, overriding the default coloring scheme imposed by the DislocationVis.coloring_mode setting. The custom color is only used if its RGB components are non-negative (i.e. in the range 0-1); otherwise the line will be rendered using the computed color depending on the line’s Burgers vector.

Default:

(-1.0, -1.0, -1.0)

property id

The unique numerical identifier of this dislocation line within the DislocationNetwork. This is simply the 0-based index of the line in the DislocationNetwork.lines list.

Important: This identifier is derived from the arbitrary storage order of the lines in the network and cannot be used to identify the same dislocation in another simulation snapshot.

property is_infinite_line

Indicates that this dislocation is an infinite line passing through a periodic simulation box boundary. A dislocation is considered infinite if it is a closed loop but its start and end do not coincide (because they are located in different periodic images).

property is_loop

Indicates whether this line forms a loop, i.e., its end is connected to its start point. Note that an infinite dislocation line passing through a periodic simulation cell boundary is also considered a logical loop (see is_infinite_line property).

property length

Computes the length of this dislocation line in simulation units of length, integrating the piece-wise linear segments it is made of.

point_along_line(t)

Returns the Cartesian coordinates of a point on the dislocation line. The location to be calculated must be specified in the form of a fractional position t along the continuous dislocation line.

Parameters:

t (float) – Normalized path coordinate in the range [0,1]

Returns:

The xyz coordinates of the requested point on the dislocation line.

New in version 3.10.0.

property points

The sequence of spatial points that define the curved shape of this dislocation (in simulation coordinates). This is a N x 3 Numpy array, with N>2 being the number of points along the line.

For true dislocation loops, the first and the last point in the list coincide exactly. For infinite lines, the first and the last point coincide modulo a periodicity vector of the simulation domain.

The point sequence always forms a continuous line, which may lead outside the primary SimulationCell if periodic boundary conditions (PBCs) are used, i.e., only the start of the dislocation is always inside the primary simulation cell but its end point may not. Thus, the line is stored in unwrapped form. A wrapping happens ad-hoc during visualization, when the DislocationVis element renders the dislocation network or if the network is exported to a file format, e.g. VTK, which does not support PBCs.

property spatial_burgers_vector

The Burgers vector of the segment, expressed in the global coordinate system of the simulation. This vector is calculated by transforming the true Burgers vector from the local lattice coordinate system to the global simulation coordinate system using the average orientation matrix of the crystal cluster the dislocation segment is embedded in.

property true_burgers_vector

The Burgers vector of the dislocation expressed in the local coordinate system of the crystal the dislocation is located in. The true Burgers vectors of two dislocation lines may only be added if both belong to the same cluster_id.

find_nodes()

Returns a list of all unique dislocation nodes in the network.

For a detailed description of what a node is, see the Connector class. The list returned by this method contains one (arbitrary) Connector from each network node, in no particular order. Each of these connectors serves as access into a node and can be used to visit the other connectors (dislocation arms) in the same node via Connector.arms:

network = data.dislocations

for node in network.find_nodes():
    print(node.arm_count, node.position)
    for arm in node.arms():
        print(arm.line.true_burgers_vector)

New in version 3.10.2.

property lines

The list of dislocation lines in this dislocation network. This list-like object contains Line objects in arbitrary order and is read-only.

set_line(index, true_burgers_vector=None, cluster_id=None, points=None, custom_color=None)

This method can be used to manipulate certain aspects of a Line in the network. Fields for which no new value is specified will keep their current values.

Parameters:
  • index – The zero-based index of the dislocation line to modify in the lines array.

  • true_burgers_vector – The new lattice-space Burgers vector (true_burgers_vector).

  • cluster_id – The numeric ID of the crystallite cluster the dislocation line is embedded in (cluster_id).

  • points – An \((N, 3)\) NumPy array of Cartesian coordinates containing the dislocation’s vertices (points).

  • custom_color – RGB color to be used for rendering the line instead of the automatically determined color (custom_color).

Example of a user-defined modifier function that manipulates the dislocation line data:

import numpy as np

def modify(frame: int, data: DataCollection):

    # Flip Burgers vector and line sense of each dislocation:
    for index, line in enumerate(data.dislocations.lines):
        data.dislocations_.set_line(index,
            true_burgers_vector = np.negative(line.true_burgers_vector),
            points = np.flipud(line.points))

    # Highlight all 1/6[121] dislocations by giving them a red color:
    for index, line in enumerate(data.dislocations.lines):
        if np.allclose(line.true_burgers_vector, (1/6, 2/6, 1/6)):
            data.dislocations_.set_line(index, custom_color=(1, 0, 0))
class ovito.data.ElementType

Base: ovito.data.DataObject

This class describes a single “type”, for example a particle or bond type, that is part of a typed property array.

The ElementType class is the generic base class used for all type descriptors in OVITO. It stores general attributes such as a type’s unique numeric id, its human-readable name, and its display color. Note that, for certain typed properties, OVITO uses more specific sub-classes such as ParticleType and BondType, which can store additional attributes such as radius and mass.

The ElementType instances associated with a typed property are found in the Property.types list.

You can use the Property.type_by_id() and Property.type_by_name() methods to look up a certain ElementType based on a numeric identifier or name string.

property color

The color used when rendering elements of this type. This is a RGB tuple with components in the range 0.0 – 1.0.

Default:

(1.0, 1.0, 1.0)

property enabled

Controls whether this type is currently active or inactive. This flag currently has a meaning only in the context of atomic structure identification. Some analysis modifiers manage a list of the structure types they can identify (e.g. FCC, BCC, etc.). The identification of individual structure types can be turned on or off by the user by changing their enabled flag. See StructureIdentificationModifier.structures for further information.

Default:

True

property id

The unique numeric identifier of the type (typically some positive int). The identifier is and must be unique among all element types in the types list of a typed Property. Thus, if you create a new element type, make sure you give it a unique id before inserting it into the types list of a typed property.

Default:

0

property name

The name of this type, e.g. the chemical element symbol of an atom type. This string may be empty, in which case its numeric id is the only way of referring to this type.

Default:

''

class ovito.data.Lines

Base: ovito.data.PropertyContainer

New in version 3.10.0.

The Lines class represents one or more 3D polylines. You can create an instance of this class in a DataCollection using the GenerateTrajectoryLinesModifier or the DataCollection.lines.create() method.

You can retrieve existing Lines objects from a pipeline’s output through the DataCollection.lines dictionary view. Each Lines object has a unique identifier name that serves as a lookup key.

Lines objects are always associated with LinesVis element, which controls the visual appearance of the lines in rendered images. You can access the visual element through the vis attribute provided by the DataObject base class. The LinesVis element provides the capability to visualize a local quantity defined at each line vertex using pseudo-color mapping.

The Lines container uses the following standard properties with predefined names and data layouts. Additional per-vertex properties may be added using the create_property() method of the base class.

Property name

Python access

Data type

Component names

Color

float32

R, G, B

Position

positions

float64

X, Y, Z

Section

sections

int64

Time

time_stamps

int32

Data model

The Lines data object type is PropertyContainer, which means it consist of a set of uniform property arrays. The standard property Position stores the vertex coordinates. When rendering a Lines object, consecutive vertices get connected by linear line segments to form a polyline (which is also called a line “section”). To denote the end of a contiguous polyline and the start of a new one, the property Section must be filled with section identifiers, which mark consecutive sequences of vertices, each forming a separate polyline. Section identifiers are abitrary integer numbers that stay the same within a polyline but change from one polyline to the next:

../_images/lines_example_definition.svg
Example: Definition of three polylines of length 2, 3, and 2, respectively.

Vertex

Position

Section

0

\((x_0, y_0, z_0)\)

0

1

\((x_1, y_1, z_1)\)

0

2

\((x_2, y_2, z_2)\)

1

3

\((x_3, y_3, z_3)\)

1

4

\((x_4, y_4, z_4)\)

1

5

\((x_5, y_5, z_5)\)

2

6

\((x_6, y_6, z_6)\)

2

The Lines class provides the create_line() method to add new polylines to the container. It appends a list of vertex coordinates to the Position property and automatically assigns a new unique Section value to these vertices to define a new polyline.

Note

In a Lines object created by the GenerateTrajectoryLinesModifier, the Section property reflects the unique identifiers of the particles that traced the individual trajectory lines.

If the Time property is present, it stores the animation frame number at which each vertex should appear in a trajectory animation. This feature allows to animate the lines over time and is used for particle trajectory visualization. The gradual rendering of lines is only active if the LinesVis.upto_current_time option is enabled in the attached visual element.

create_line(positions)

Adds a new section (a polyline) to an existing Lines container. The container’s vertex count will be incremented by the number of newly inserted points. The method copies positions into the Position property array after extending the array and gives the new polyline a unique Section property value.

Parameters:

positions (array-like) – The xyz coordinates for the new lines section (N x 3 array). N must be at least 2.

Returns:

The unique section identifier that was assigned to the newly added polyline.

Return type:

int

property positions

The Property array containing the XYZ coordinates of the line vertices (standard property Position). May be None if the property is not defined yet. Use create_property() to add the property to the container if necessary. Use positions_ (with an underscore) to access an independent copy of the array, whose contents can be safely modified in place.

property sections

The Property array with the section each line vertex belongs to (standard property Section). May be None if the property is not defined yet. Use create_property() to add the property to the container if necessary. Use sections_ (with an underscore) to access an independent copy of the array, whose contents can be safely modified in place.

property time_stamps

The Property array with the time stamps of the line vertices (standard property Time). May be None if the property is not defined yet. Use create_property() to add the property to the container if necessary. Use time_stamps_ (with an underscore) to access an independent copy of the array, whose contents can be safely modified in place.

class ovito.data.NearestNeighborFinder(N, data_collection)

A utility class that finds the N nearest neighbors of a particle.

See also

To find all neighbors within a spherical cutoff region around another particle, use CutoffNeighborFinder instead.

The constructor takes the requested number of nearest neighbors, N, and a DataCollection containing the input particles and the optional simulation cell. N must be a positive integer not greater than 64, which is the built-in maximum of neighbors supported by this class.

Note

Keep in mind that, if the system contains only N particles or less, and if the simulation does not use periodic boundary conditions, then the neighbor finder will return less than the requested number of nearest neighbors.

Once the NearestNeighborFinder has been initialized, you can call its find() method to iterate over the sorted list of nearest neighbors of a given central particle:

# Set up a neighbor finder for visiting the 12 closest neighbors of each particle.
finder = NearestNeighborFinder(12, data)

# Loop over all input particles:
for index in range(data.particles.count):
    print("Nearest neighbors of particle %i:" % index)
    # Iterate over the neighbors of the current particle, starting with the closest:
    for neigh in finder.find(index):
        print(neigh.index, neigh.distance, neigh.delta)
        # The index can be used to access properties of the current neighbor, e.g.
        type_of_neighbor = data.particles.particle_types[neigh.index]

In addition, the class provides the find_at() method, which determines the N nearest particles around some arbitrary spatial location:

# Visit particles closest to some spatial point (x,y,z):
xyz_coords = (0.0, 0.0, 0.0)
for neigh in finder.find_at(xyz_coords):
    print(neigh.index, neigh.distance, neigh.delta)
find(index)

Returns an iterator that visits the N nearest neighbors of the given particle in order of ascending distance.

Parameters:

index (int) – The zero-based index of the central particle whose neighbors should be determined.

Returns:

A Python iterator that visits the N nearest neighbors of the central particle in order of ascending distance. For each neighbor being visited, the iterator returns an object having the following attributes:

  • index: The global index of the current neighbor particle.

  • distance: The distance of the current neighbor from the central particle.

  • distance_squared: The squared neighbor distance.

  • delta: The three-dimensional vector connecting the central particle with the current neighbor (correctly taking into account periodic boundary conditions).

The index can be used to look up properties of the neighbor particle, as demonstrated in the first example code above.

Note that several periodic images of the same particle may be visited if the periodic simulation cell is sufficiently small. Then the same particle index will appear more than once in the neighbor list. In fact, the central particle may be among its own neighbors in a sufficiently small periodic simulation cell. However, the computed neighbor vector (delta) will be unique for each image of a neighboring particle.

The number of neighbors actually visited may be smaller than the requested number, N, if the system contains too few particles and is non-periodic.

Note that the find() method will not find other particles located exactly at the same spatial position as the central particle for technical reasons. To find such particles too, which are positioned exactly on top of each other, use find_at() instead.

find_all(indices=None)

Finds the N nearest neighbors of each particle in the system or of the subset of particles specified by indices. This is the batch-processing version of find(), allowing you to efficiently compute the neighbor lists and neighbor vectors of several particles at once, without explicit for-loop and by making use of all parallel processor cores.

The method returns two NumPy arrays:

neigh_idx : NumPy array of shape (M, N) storing the indices of neighbor particles, with M equal to len(indices) or, if indices is None, the total number of particles in the system. N refers to the number of nearest neighbors requested in the NearestNeighborFinder constructor. The computed indices in this array can be used to look up properties of neighbor particles in the global Particles object.

neigh_vec : NumPy array of shape (M, N, 3) storing the xyz components of the three-dimensional neighbor vectors (“delta”), which connect the M central particles with their N respective nearest neighbors.

Parameters:

indices – List of zero-based particle indices for which the neighbor lists should be computed. If left unspecified, neighbor lists will be computed for every particle in the system.

Returns:

(neigh_idx, neigh_vec)

Tip

To compute all pair-wise distances in one go, i.e. the 2-norms of the neighbor vectors, you can do:

distances = numpy.linalg.norm(neigh_vec, axis=2)   # Yields (M,N) array of neighbor distances
find_at(coords)

Returns an iterator that visits the N nearest particles around a spatial point given by coords in order of ascending distance. Unlike the find() method, which queries the nearest neighbors of a physical particle, find_at() allows searching for nearby particles at arbitrary locations in space.

Parameters:

coords – A coordinate triplet (x,y,z) specifying the spatial location where the N nearest particles should be queried.

Returns:

A Python iterator that visits the N nearest neighbors in order of ascending distance. For each visited particle the iterator returns an object with the following attributes:

  • index: The index of the current particle (starting at 0).

  • distance: The distance of the current neighbor from the query location.

  • distance_squared: The squared distance to the query location.

  • delta: The three-dimensional vector from the query point to the current particle (correctly taking into account periodic boundary conditions).

If there is a particle located exactly at the query location coords, it will be among the returned neighbors. This is in contrast to the find() function, which skips the central particle itself.

The number of neighbors actually visited may be smaller than the requested number, N, if the system contains too few particles and is non-periodic.

class ovito.data.ParticleType

Base: ovito.data.ElementType

This data object describes one particle or atom type. In atomistic simulations, each chemical element is typically represented by an instance of the ParticleType class. The attributes of this class control how the particles of that type get visualized in terms of e.g. color, particle radius, shape, etc.

The ParticleType class inherits several general attribute fields from its base class ElementType, e.g. the color, name and id fields. It adds specific fields for particles: radius and shape. Furthermore, the class has additional fields controlling the visual appearance of particles with user-defined shapes.

The ParticleType instances all live in the Property.types list of the 'Particle Type' standard Property array, which is accessible as Particles.particle_types. The association of particles with particle types is established through the unique type IDs. The following code shows how to first list all unique particle types defined for a structure and then print the each particle’s type ID by iterating over the data array:

# Access the property with the name 'Particle Type':
type_property = data.particles.particle_types

# Print list of particle types (their numeric IDs and names)
for t in type_property.types:
    print(f'ID {t.id} -> {t.name}')

# Print the numeric type ID of each particle:
for tid in type_property[...]:
    print(tid)

A common task is to look up the ParticleType that corresponds to a given numeric type ID. For this, the Property class provides the type_by_id() method:

# Look up the particle type with unique ID 2:
t = type_property.type_by_id(2)
print(t.name, t.color, t.radius)

# Iterate over all particles and print their type's name:
for tid in type_property[...]:
    print(type_property.type_by_id(tid).name)

Another common operation is to look up a particle type by name, for example the type representing a certain chemical element. For this type of look-up the type_by_name() method is available, which assumes that types have unique names:

print(type_property.type_by_name('Si').id)   # Print numeric ID of type 'Si'
property backface_culling

Activates back-face culling for the user-defined particle shape mesh to speed up rendering. If turned on, polygonal sides of the shape mesh facing away from the viewer will not be rendered. You can turn this option off if the particle’s shape is not closed and two-sided rendering is required. This option only has an effect if a user-defined shape has been assigned to the particle type using the load_shape() method.

Default:

True

property highlight_edges

Activates the highlighting of the polygonal edges of the user-defined particle shape during rendering. This option only has an effect if a user-defined shape has been assigned to the particle type using the load_shape() method.

Default:

False

load_defaults()

Given the type’s chemical name, which must have been set before invoking this method, initializes the type’s color, radius, vdw_radius and , mass fields with default values from OVITO’s internal database of chemical elements.

load_shape(filepath: str)

Assigns a user-defined shape to the particle type. Particles of this type will subsequently be rendered using the polyhedral :py:mesh loaded from the given file. The method will automatically detect the format of the geometry file and supports standard file formats such as OBJ, STL and VTK that contain triangle meshes, see this table.

The shape loaded from the geometry file will be scaled with the radius value set for this particle type or the per-particle value stored in the Radius particle property if present. The shape of each particle will be rendered such that its origin is located at the coordinates of the particle (Position property).

The following example script demonstrates how to load a user-defined shape for the first particle type (index 0) loaded from a LAMMPS dump file, which can be accessed through the Property.types list of the Particle Type particle property.

pipeline = import_file("input/simulation.dump")
pipeline.add_to_scene()

types = pipeline.source.data.particles_.particle_types_
types.type_by_id_(1).load_shape("input/tetrahedron.vtk")
types.type_by_id_(1).highlight_edges = True
property mass

The mass of this particle type.

Default:

0.0

property mesh

The TriangleMesh object to be used as custom shape for rendering particles of this type. You can either programmatically create a TriangleMesh from a list of vertices and faces and assign it to this field, or let the load_shape() method read the shape mesh from a separate geometry file. Also some file readers (e.g. GSD and Aspherix) may generate the shape mesh automatically based on information found in the simulation file.

The ParticlesVis element, which is responsible for visualizing a particle system, will render an instance of the mesh at each particle site, uniformely scaled by the particle’s radius, translated by the coordinates taken from the Position particle property, and rotated by the quaternion transformation taken from the Orientation particle property.

Note: This mesh will be ignored by the ParticlesVis element unless the type’s shape is set to ParticlesVis.Shape.Mesh.

Default:

None

New in version 3.8.0.

property radius

This attribute controls the display radius of all particles of this type.

When set to zero, particles of this type will be rendered using the standard size specified by the ParticlesVis.radius parameter. Furthermore, precedence is given to any per-particle sizes assigned to the Radius particle property if that property has been defined.

Default:

0.0

The following example script demonstrates how to set the display radii of two particle types loaded from a simulation file, which can be accessed through the Property.types list of the Particle Type particle property.

pipeline = import_file("input/simulation.dump")
pipeline.add_to_scene()

def setup_particle_types(frame, data):
    types = data.particles_.particle_types_
    types.type_by_id_(1).name = "Cu"
    types.type_by_id_(1).radius = 1.35
    types.type_by_id_(2).name = "Zr"
    types.type_by_id_(2).radius = 1.55
pipeline.modifiers.append(setup_particle_types)
property shape

Selects the geometric shape used when rendering particles of this type. Supported modes are:

  • ParticlesVis.Shape.Unspecified (default)

  • ParticlesVis.Shape.Sphere

  • ParticlesVis.Shape.Box

  • ParticlesVis.Shape.Circle

  • ParticlesVis.Shape.Square

  • ParticlesVis.Shape.Cylinder

  • ParticlesVis.Shape.Spherocylinder

  • ParticlesVis.Shape.Mesh

By default, the standard particle shape that is set in the ParticlesVis visual element is used to render particles of this type. Parameter values other than Unspecified allow you to control the rendering shape on a per-type basis. Mode Sphere includes ellipsoid and superquadric particle shapes, which are enabled by the presence of the Aspherical Shape and Superquadric Roundness particle properties.

The load_shape() method lets you specify a user-defined mesh geometry for this particle type. Calling this method automatically switches the shape parameter to mode Mesh.

Setting the shapes of particle types permanently, i.e., for all frames of a loaded simulation trajectory, typically requires a user-defined modifier function. This function is inserted into the Pipeline to make the necessary changes to the ParticleType objects associated with the Property named Particle Type:

from ovito.io import import_file
from ovito.vis import *

# Load a simulation file containing numeric particle types 1, 2, 3, ...
pipeline = import_file("input/nylon.data")
pipeline.add_to_scene()

# Set the default particle shape in the ParticlesVis visual element, 
# which will be used by all particle types for which we do not specify a different shape below.
pipeline.compute().particles.vis.shape = ParticlesVis.Shape.Box
pipeline.compute().particles.vis.radius = 1.0

# A user-defined modifier function that configures the shapes of particle types 1 and 2:
def setup_particle_types(frame, data): 
    # Write access to property 'Particle Type':
    types = data.particles_.particle_types_  
    # Write access to numeric ParticleTypes, which are sub-objects of the Property object:
    types.type_by_id_(1).radius = 0.5
    types.type_by_id_(1).shape = ParticlesVis.Shape.Cylinder
    types.type_by_id_(2).radius = 1.2
    types.type_by_id_(2).shape = ParticlesVis.Shape.Sphere
pipeline.modifiers.append(setup_particle_types)

# Render a picture of the 3d scene:
vp = Viewport(camera_dir = (-2,1,-1))
vp.zoom_all()
vp.render_image(filename='output/particles.png', size=(320,240), renderer=TachyonRenderer())
property use_mesh_color

Use the intrinsic mesh color(s) instead of the particle color when rendering particles of this type. This option only has an effect if a user-defined shape :py:mesh has been assigned to this particle type, e.g., by calling the load_shape() method.

Default:

False

property vdw_radius

The van der Waals radius of the particle type. This value is used by the CreateBondsModifier to decide which pairs of particles are close enough to be connected by a bond. In contrast to the radius parameter, the van der Waals radius does not affect the visual appearance of the particles of this type.

Default:

0.0

class ovito.data.Particles

Base: ovito.data.PropertyContainer

This object stores a system of particles and their properties. Additional things which are typically associated with molecular systems, e.g. bonds, angles, etc. are stored in corresponding sub-objects.

A Particles object is usually part of a DataCollection where it can be found via the DataCollection.particles property.

The total number of particles is specified by the count attribute, which the Particles class inherits from its PropertyContainer base class.

Particles are usually associated with a set of properties, e.g. position, type, velocity. Each of the properties is represented by a separate Property data object, which is basically an array of numeric values, one for each particle in the system. A particle property is identified by its unique name and can be looked up via the dictionary interface of the PropertyContainer base class. OVITO predefines a set of standard properties, which have a fixed data layout, meaning, and role:

Standard property name

Data type

Component names

Angular Momentum

float64

X, Y, Z

Angular Velocity

float64

X, Y, Z

Aspherical Shape

float32

X, Y, Z

Centrosymmetry

float64

Charge

float64

Cluster

int64

Color

float32

R, G, B

Coordination

int32

Deformation Gradient

float64

XX, YX, ZX, XY, YY, ZY, XZ, YZ, ZZ

Dipole Magnitude

float64

Dipole Orientation

float64

X, Y, Z

Displacement Magnitude

float64

Displacement

float64

X, Y, Z

DNA Strand

int32

Elastic Deformation Gradient

float64

XX, YX, ZX, XY, YY, ZY, XZ, YZ, ZZ

Elastic Strain

float64

XX, YY, ZZ, XY, XZ, YZ

Force

float64

X, Y, Z

Kinetic Energy

float64

Mass

float64

Molecule Identifier

int64

Molecule Type

int32

Nucleobase

int32

Nucleotide Axis

float64

X, Y, Z

Nucleotide Normal

float64

X, Y, Z

Orientation

float32

X, Y, Z, W

Particle Identifier

int64

Particle Type

int32

Periodic Image

int32

X, Y, Z

Position

float64

X, Y, Z

Potential Energy

float64

Radius

float32

Rotation

float64

X, Y, Z, W

Selection

int8

Spin

float64

Strain Tensor

float64

XX, YY, ZZ, XY, XZ, YZ

Stress Tensor

float64

XX, YY, ZZ, XY, XZ, YZ

Stretch Tensor

float64

XX, YY, ZZ, XY, XZ, YZ

Structure Type

int32

Superquadric Roundness

float32

Phi, Theta

Torque

float64

X, Y, Z

Total Energy

float64

Transparency

float32

Vector Color

float32

R, G, B

Velocity Magnitude

float64

Velocity

float64

X, Y, Z

For some of the most important properties, this container class provides quick access getters such as positions, identifiers, or particle_types to look them up:

coords = data.particles.positions

User-defined particle properties having non-standard names, and standard properties for which no quick access getter exists, can be looked up by literal name:

mol_ids = data.particles['Molecule Identifier']

For more information on how to add or modify particle properties, please see the PropertyContainer and Property classes.

add_particle(position)

Adds a new particle to the model. The particle count will be incremented by one. The method assigns position to the Position property of the new particle. The values of all other properties are initialized to zero.

Parameters:

position (array-like) – The xyz coordinates for the new particle.

Returns:

The index of the newly created particle, i.e. (Particles.count-1).

property angles

A PropertyContainer storing the list of angles defined for the molecular model (may be None).

property bonds

The Bonds data object storing the list of bonds and their properties (may be None).

property colors

The Property data array for the Color standard particle property; or None if that property is undefined.

create_bonds(vis_params=None, **params)

This convenience method conditionally creates and associates a Bonds object with this Particles parent object. If there is already an existing bonds object (bonds is not None), then that bonds object is replaced with a modifiable copy if necessary. The attached BondsVis element is preserved.

Parameters:
  • params – Key/value pairs passed to the method as keyword arguments are used to set attributes of the Bonds object (even if the bonds object already existed).

  • vis_params (Mapping[str, Any]) – Optional dictionary to initialize attributes of the attached BondsVis element (only used if the bonds object is newly created by the method).

Return type:

ovito.data.Bonds

The logic of this method is roughly equivalent to the following code:

def create_bonds(particles: Particles, vis_params=None, **params) -> Bonds:
    if particles.bonds is None:
        particles.bonds = Bonds()
        if vis_params:
            for name, value in vis_params.items(): setattr(particles.bonds.vis, name, value)
    for name, value in params.items(): setattr(particles.bonds_, name, value)
    return particles.bonds_

Usage example:

pairs = [(0, 1), (1, 2), (2, 0)] # Pairs of particle indices to connect by bonds
bonds = data.particles_.create_bonds(count=len(pairs), vis_params={'width': 0.6})
bonds.create_property('Topology', data=pairs)

New in version 3.7.4.

delta_vector(a, b, cell, return_pbcvec=False)

Computes the vector connecting two particles a and b in a periodic simulation cell by applying the minimum image convention.

This is a convenience wrapper for the SimulationCell.delta_vector() method, which computes the vector between two arbitrary spatial locations \(r_a\) and \(r_b\) taking into account periodic boundary conditions. The version of the method described here takes two particle indices a and b as input, computing the shortest vector \({\Delta} = (r_b - r_a)\) between them using the minimum image convention. Please see the SimulationCell.delta_vector() method for further information.

Parameters:
  • a – Zero-based index of the first input particle. This may also be an array of particle indices.

  • b – Zero-based index of the second input particle. This may also be an array of particle indices with the same length as a.

  • cell (SimulationCell) – The periodic domain. Typically, DataCollection.cell is used as argument here.

  • return_pbcvec (bool) – If True, also returns the vector \(n\), which specifies how often the computed particle-to-particle vector crosses the cell’s face.

Returns:

The delta vector and, optionally, the vector \(n\).

property dihedrals

A PropertyContainer storing the list of dihedrals defined for the molecular model (may be None).

property forces

The Property data array for the Force standard particle property; or None if that property is undefined.

property identifiers

Returns a Property data array containing the values of the Particle Identifier standard particle property; or None if that particle property does not exist.

The property array stores the numerical IDs that are typically used by simulation codes to uniquely identify individual particles.

Note

A particle identifier is an arbitrary and unique 64-bit integer value permanently associated with a particle. In contrast, the particle’s index is implicitly given by the particle’s current position within the particles list.

If you delete some of the particles from the system, using the DeleteSelectedModifier or the delete_elements() method for example, the indices of the remaining particles get typically shifted but their unique IDs stay the same.

Some of OVITO’s simulation file readers provide the option to sort the list of particles by ID during import to obtain a stable ordering. Generally, however, the storage order of particles is arbitrary and can vary between frames of a trajectory. The remap_indices() method can be useful in this situation.

Note

The value of identifiers may be None, which means particles don’t have any identifiers. Many operations in OVITO then assume that the ordering and total count of particles are constant throughout the entire simulation trajectory and the identities are implicitly given by the particles’ indices.

Given some list of zero-based particle indices, determining the corresponding unique identifiers requires just a simple NumPy indexing expression:

query_indices = [0, 7, 3, 2] # <-- zero-based particle indices for which we want to look up IDs
ids = data.particles.identifiers[query_indices]
assert len(ids) == len(query_indices)

The reverse lookup, i.e., finding the indices at which particles with certain IDs are stored in the list, requires some more work. That’s because particles may be stored in arbitrary order, i.e., the sequence of unique IDs is generally not sorted nor contigous. A rather slow approach is to search through the entire array of IDs to locate the one we are looking for:

query_id = 37 # <-- a unique particle ID we are looking for
index = numpy.argwhere(data.particles.identifiers == query_id)[0,0]
assert data.particles.identifiers[index] == query_id

We can speed things up with some extra effort, which pays off when there is a need to look up several particle IDs. To this end, we first sort the list of IDs, then perform a more efficient sorted search, and finally map the found indices back to the original particle ordering:

query_ids = [2, 37, 8] # <-- some unique particle IDs we want to look up all at once
ordering = numpy.argsort(data.particles.identifiers)
indices = ordering[numpy.searchsorted(data.particles.identifiers, query_ids, sorter=ordering)]
assert numpy.array_equal(data.particles.identifiers[indices], query_ids)

The above assert statements are for illustration purposes only.

property impropers

A PropertyContainer storing the list of impropers defined for the molecular model (may be None).

property masses

The Property data array for the Mass standard particle property; or None if that property is undefined.

property orientations

The Property data array for the Orientation standard particle property; or None if that property is undefined.

property particle_types

The Property data array for the Particle Type standard particle property; or None if that property is undefined.

property positions

Returns the Property data array storing the particle coordinates, i.e. the values of the Position standard particle property. Accessing this field is equivalent to a name-based lookup in the PropertyContainer:

assert data.particles.positions is (data.particles['Position'] if 'Position' in data.particles else None)

Under special circumstances the Position particle property might not be defined (yet). Then the value is None.

Note

The returned Property will likely be write-protected. If you intend to modify (some of) the particle coordinates in the property array, request a modifiable version of the array by using the underscore notation:

data.particles_.positions_[:] = new_coordinates

Alternatively, you can use the create_property() method to newly create or overwrite the entire property:

data.particles_.create_property('Position', data=new_coordinates)
remap_indices(particles: Particles, indices: Sequence[int] = None) numpy.ndarray | slice

In case the storage order of atoms or particles changes during the course of a simulation, this method can determine the mapping of particles from one snapshot of the trajectory to another. It uses the unique identifiers of the particles to do that.

Given two data collections A and B containing the same set of particles but in different order, remap_indices() determines for each particle in B the zero-based index at which the same particle is found in A. For instance:

>>> A = pipeline.compute(frame=0)
>>> B = pipeline.compute(frame=1)
>>> A.particles.identifiers[...]
[8 101  5 30 99]
>>> B.particles.identifiers[...]
[5 101 30 99  8]
>>> A.particles.remap_indices(B.particles)
[2 1 3 4 0]

The index mapping generated by remap_indices() allows you to retrieve property values of particles in A in the same order in which they appear in B, making it easy to perform computations involving property values at different trajectory timesteps, e.g.:

mapping = A.particles.remap_indices(B.particles)
displacements = B.particles.positions - A.particles.positions[mapping]

remap_indices() compares the unique identifiers stored in the Particle Identifier property arrays of both snapshots to compute the index permutation map. If this property is not defined, which may be the case if the imported trajectory file did not contain atom IDs, the remap_indices() method simply assumes that both snapshots use the same constant storage order and returns the identity mapping - as a Python slice object for optimal performance when being used for NumPy indexing. A slice object is also returned in case the ordering of particle IDs turns out to be the same in both snapshots and no remapping is necessary.

Note

An error will be raised if particles with duplicate IDs occur in snapshot A - but it is okay if B contains duplicate IDs. Furthermore, it is not an error if A contains additional particles that are not present in B - as long as all particles from B are found in A.

The default behavior of the method is to look up all particles of B in A. But the index mapping can also be established just for a subset of particles from B by supplying the optional parameter indices. The method expects an array of zero-based indices specifying which particles from snapshot B should be looked up in snapshot A. The returned mapping will have the same length as indices. Example:

# The numeric ID of atom type 'H':
hydrogen_type = B.particles.particle_types.type_by_name('H').id
# Determine the indices of all H atoms in data collection B:
hydrogen_indices = numpy.flatnonzero(B.particles.particle_types == hydrogen_type)

# Determine the corresponding indices of the same atoms in data collection A:
mapping = A.particles.remap_indices(B.particles, hydrogen_indices)

# In snapshot A the same particles are all H atoms too:
assert numpy.all(A.particles.particle_types[mapping] == hydrogen_type)

New in version 3.7.5.

property selection

The Property data array for the Selection standard particle property; or None if that property is undefined.

property structure_types

The Property data array for the Structure Type standard particle property; or None if that property is undefined.

property velocities

The Property data array for the Velocity standard particle property; or None if that property is undefined.

class ovito.data.Property

Base: ovito.data.DataObject

A storage array for the values of one uniform property of particles, bonds, voxel grid cells, etc. For example, the “Position” property of particles is represented by one Property object storing the xyz cordinates of all the particles.

Property objects are always managed by a specific sub-type of the PropertyContainer class, for example Particles, Bonds, VoxelGrid, or DataTable. These container classes allow to add and remove properties to the data elements they represent. The properties within the PropertyContainer are accessed by name. Here, for example, the particle property holding the particle coordinates:

positions = data.particles['Position']

This Property object behaves almost like a regular NumPy array. For example, you can access the value for the i-th element using array indexing:

print('XYZ coordinates of first particle:', positions[0])
print('z-coordinate of sixth particle:', positions[5,2])
print(positions.shape)   # --> (data.particles.count, 3)

Since the “Positionstandard property has three components, this Property object is an array of shape (N,3). Properties can be either vectorial or scalar, and they can hold uniform data types float64, float32, int8, int32 or int64.

If you want to set or modify the values stored in a property array, make sure you are working with a modifiable version of the Property object by employing the underscore notation, e.g.:

modifiable_positions = data.particles_['Position_']
modifiable_positions[0] += (2.0, 0.0, 0.5)

Typed properties

In OVITO, the standard particle propertyParticle Type” contains the types of particles encoded as integer values, e.g.:

>>> data = pipeline.compute()
>>> type_property = data.particles['Particle Type']
>>> print(type_property[...])
[2 1 3 ..., 2 1 2]

The property array stores numeric type identifiers denoting each particle’s chemical type (e.g. 1=Cu, 2=Ni, 3=Fe, etc.). The mapping of unique numeric IDs to human-readable type names is found in the types list, which is attached to the Property array, making it a so-called typed property. This list contains one ParticleType descriptor per unique numerical type, specifying its human-readable name as well as other attributes such as display color, radius and mass:

>>> for t in type_property.types:
...     print(t.id, t.name, t.color, t.radius)
...
1 Cu [0.188 0.313 0.972] 0.74
2 Ni [0.564 0.564 0.564] 0.77
3 Fe [1 0.050 0.050] 0.74

Numeric type IDs typically start at 1 and form a consecutive sequence as in the example above. But they don’t have to. The descriptors may be listed in any order and their numeric IDs may be arbitrary integers. Thus, in general, it is not valid to directly use a numeric type ID as an index into the types list. Instead, the type_by_id() method should be used to look up the ParticleType descriptor for a given numeric ID:

>>> for i,t in enumerate(type_property): # loop over the type IDs of particles
...     print(f"Atom {i} is of type {type_property.type_by_id(t).name}")
...
Atom 0 is of type Ni
Atom 1 is of type Cu
Atom 2 is of type Fe
Atom 3 is of type Cu

Similarly, a type_by_name() method exists that allows to look up a ParticleType from the types list by name. For example, to count the number of Fe atoms in a system, we first need to determine the numeric ID of the type “Fe” and then count the number of occurences of the value in the Property array:

>>> Fe_type_id = type_property.type_by_name('Fe').id  # Determine numeric ID of the 'Fe' type.
>>> numpy.count_nonzero(type_property == Fe_type_id)  # Count particles having that type ID.
957

Note that the data model supports multiple type classifications per particle. For example, while the “Particle Typestandard particle property, stores the chemical types of atoms (e.g. C, H, Fe, …), the “Structure Type” property stores the structural lattice types computed for each atom (e.g. FCC, BCC, …). In other words, multiple typed properties can co-exist to define several orthogonal classifications, and each typed property maintains its separate list of type descriptors in the associated types list.

New types can be added to a typed property either using add_type_id() or add_type_name(). Use the former method if it is important that the new type gets a particular numeric ID (which must not collide with existing types in the types list). Use the latter method if you don’t care about the numeric ID and let the method automatically assign a unique ID to the new type.

See also

Typed properties

add_type_id(id: int, container: PropertyContainer, name: str = '') ElementType | ParticleType | BondType

Creates a new numeric ElementType with the given numeric id and an optional human-readable name and adds it to this property’s types list. If the list already contains an existing element type with the same numeric id, that existing type will be returned (without updating its name).

Additionally, you must specify the PropertyContainer containing this property object as second parameter, because it determines the kind of ElementType to create. For example, when calling add_type_id() on the property “Particle Type” of a Particles container, this method will create a new ParticleType object – a specific sub-class of the more general ElementType class. Furthermore, if name matches one of the standard type names predefined for that particle property, e.g., a chemical symbol in case of the “Particle Type” property, the type’s display color, radius, and mass will be preconfigured (as if load_defaults() was called).

type_property = data.particles_.create_property("Particle Type")
type_1 = type_property.add_type_id(1, data.particles, name="A")
type_2 = type_property.add_type_id(2, data.particles, name="B")

# Configure visual appearance of the two ParticleTypes
type_1.radius = 0.9; type_1.color = (1.0, 0.0, 0.0)
type_2.radius = 1.2; type_2.color = (0.0, 0.0, 1.0)

# Randomly assign types "A" (1) or "B" (2) to the particles
type_property[...] = numpy.random.randint(low=1, high=1+2, size=data.particles.count)

See also

add_type_name()

New in version 3.9.0.

add_type_name(name: str, container: PropertyContainer) ElementType | ParticleType | BondType

Creates a new ElementType with the given human-readable name and adds it to this property’s types list. A unique numeric id will be automatically assigned to the type (starting at 1). If the list already contains an existing element type of the same name, that existing type will be returned.

Additionally, you must specify the PropertyContainer containing this property object as second parameter, because it determines the kind of ElementType to create. For example, when calling add_type_name() on the property “Particle Type” of a Particles container, this method will create a new ParticleType object – a specific sub-class of the more general ElementType class. Furthermore, if name matches one of the standard type names predefined for that particle property, e.g., a chemical symbol in case of the “Particle Type” property, the type’s display color, radius, and mass will be preconfigured (as if load_defaults() was called).

type_property = data.particles_.create_property("Particle Type")
type_Au = type_property.add_type_name("Au", data.particles)
type_Ag = type_property.add_type_name("Ag", data.particles)

# Randomly assign types "Au" or "Ag" to the particles
type_property[...] = numpy.random.choice([type_Au.id, type_Ag.id], size=data.particles.count)

New in version 3.9.0.

property component_count

The number of vector components if this is a vector property; or 1 if this is a scalar property.

property component_names

The list of component names if this is a vectorial property. For example, the Position particle property has three components: ['X', 'Y', 'Z'].

The number of elements in this list must always be equal to component_count or zero, in which case the property components are referenced by numeric index (1, 2, 3, …).

For predefined standard properties, OVITO automatically initializes the components list.

property name

The name of the property – a non-empty string.

The name may contain spaces, digits, or special characters, but no dots, because . is used in OVITO as a delimiter for vector component_names.

type_by_id(id, raise_error=True)

Looks up and returns the ElementType with the given unique numeric ID in this property’s types list. Depending on the parameter raise_error, raises a KeyError or returns None if no type with the numeric ID exists.

Usage example:

# Iterate over the numeric per-particle types stored in the 'Structure Type'
# particle property array and print the corresponding human-readable type names:
for index, type_id in enumerate(data.particles.structure_types):
    print("Atom {} is a {} atom".format(
        index, data.particles.structure_types.type_by_id(type_id).name))

An “underscore” version of the method exists, which should be used whenever you intend to modify the returned type object. type_by_id_() implicitly calls make_mutable() on the ElementType to make sure it can be changed without unexpected side effects:

# Give some names to the numeric atom types from a LAMMPS simulation:
data.particles_.particle_types_.type_by_id_(1).name = 'C'
data.particles_.particle_types_.type_by_id_(2).name = 'H'
type_by_name(name, raise_error=True)

Looks up and returns the ElementType with the given name in this property’s types list. If multiple types exists with the same name, the first type is returned. Depending on the parameter raise_error, raises a KeyError or returns None if there isn’t a type with that name.

Usage example:

# Look up the numeric ID of atom type Si and count how many times it appears in the 'Particle Type' array
id_Si = data.particles.particle_types.type_by_name('Si').id
Si_atom_count = numpy.count_nonzero(data.particles.particle_types == id_Si)

An “underscore” version of the method exists, which should be used whenever you intend to modify the returned type object. type_by_name_() implicitly calls make_mutable() on the ElementType to make sure it can be changed without unexpected side effects:

# Rename a structure type created by the PTM modifier:
data.particles_.structure_types_.type_by_name_('Hexagonal diamond').name = 'Wurtzite'
property types

The list of ElementType descriptors if this property is a typed property.

A typed property, such as the “Particle Type” property for particles, stores each particle’s type information, for example its chemical type, in the form of a numeric type ID in the uniform data array. The types list represents a look-up table containing a descriptor for each unique numeric type, which maps the numeric IDs to corresponding human-readable type names, chemical element symbols for example.

The types list consists of instances of the ElementType class or one of its sub-classes, describing the attributes of individual types. For each unique type, the field ElementType.id specifies its numeric identifier and ElementType.name its human-readable name. Additionally, ElementType.color specifies the color used for rendering elements of this type.

Certain typed properties in OVITO use a sub-class of ElementType to associate additional information with each unique type. For example, the “Particle Type” property array of Particles uses the ParticleType class, which defines additional fields ParticleType.radius and ParticleType.mass for each type.

Note

The type descriptors are stored in arbitrary order in the types list. Therefore, you should never use a numeric type ID as an index into this list to look up the corresponding type descriptor. Instead, use the type_by_id() method to quickly find the ElementType corresponding to a given numeric type ID.

If you want to manipulate the descriptors in the types list one by one, you should iterate over the “underscore” version of the list, types_, to automatically make all the type descriptors mutable. This shortcut implicitly invokes make_mutable() on each ElementType, making sure it can be modified without unexpected side effects:

# Get a mutable reference to the "Particle Type" property array.
type_property = data.particles_.particle_types_

# Iterate over all defined ParticleType descriptors and modify their radius.
for t in type_property.types_:
    t.radius *= 0.5

See also

class ovito.data.PropertyContainer

Base: ovito.data.DataObject

A dictionary-like object storing a set of Property objects.

It implements the collections.abc.Mapping interface. That means it can be used like a standard read-only Python dict object to access the properties by name, e.g.:

data = pipeline.compute()

positions = data.particles['Position']
has_selection = 'Selection' in data.particles
name_list = data.particles.keys()

New properties are typically added to a container with a call to create_property() as described here. To remove an existing property from a container, you can use Python’s del statement:

del data.particles_['Selection']

OVITO has several concrete implementations of the abstract PropertyContainer interface:

property count

The number of data elements in this container, e.g. the number of particles. This value is always equal to the lengths of all Property arrays managed by this container.

create_property(name, dtype=None, components=None, data=None)

Adds a new property with the given name to the container and optionally initializes its element-wise values with data. If a property with the given name already exists in the container, that existing property is returned (after replacing its contents with data if provided).

You can create standard and user-defined properties in a container. A standard property with a prescribed data layout is automatically created if name matches one of the predefined names for the container type:

The length of the provided data array must match the number of elements in the container, which is given by PropertyContainer.count. If the property to be created is a vectorial property (having \(M>1\) components), the initial data array should be of shape \((N,M)\) if provided:

colors = numpy.random.random_sample(size=(data.particles.count, 3))
data.particles_.create_property('Color', data=colors)

In general, however, data may be any value that is broadcastable to the array dimensions of the standard property (e.g. a uniform value).

If you don’t specify the function argument data, OVITO will automatically initialize the property elements with sensible default values (0 in most cases). Subsequently, you can set the property values for all or some of the elements:

prop = data.particles_.create_property('Color')
prop[...] = numpy.random.random_sample(size=prop.shape)

To create a user-defined property, specify a non-standard property name:

values = numpy.arange(0, data.particles.count, dtype=int)
data.particles_.create_property('My Integer Property', data=values)

In this case, the data type and the number of vector components of the new property are inferred from the provided NumPy array. Specifying a one-dimensional array creates a scalar property whereas a two-dimensional array creates a vectorial property. Alternatively, the dtype and components parameters can be specified explicitly if you are going to set the property values at a later time:

prop = data.particles_.create_property('My Vector Property', dtype=float, components=3)
prop[...] = numpy.random.random_sample(size = prop.shape)

If the property to be created already exists in the container, it gets replaced with a modifiable copy if necessary. The values of the property get overwritten with data in this case.

Note: If you’re creating new PropertyContainer, its element count is 0 initially. In this state the create_property() method allows you to initialize the count while adding the very first property by providing a data array of the desired length:

# An empty Particles container to begin with:
particles = Particles()

# Create 10 particles with random xyz coordinates:
xyz = numpy.random.random_sample(size=(10,3))
particles.create_property('Position', data=xyz)
assert particles.count == len(xyz)

All properties subsequently added to the container must have the same length.

Parameters:
  • name (str) – Name of the property to create.

  • data – Optional array with initial values for the new property. The size of the array must match the element count of the container and the shape must be consistent with the number of components of the property to be created.

  • dtype – Data type of the user-defined property. Must be int, float, numpy.int8, numpy.int32, numpy.int64, numpy.float32, or numpy.float64.

  • components (int) – Number of vector components of the user-defined property (1 if not specified).

Returns:

The new Property object.

delete_elements(mask)

Deletes a subset of the elements from this container. The elements to be deleted must be specified in terms of a 1-dimensional mask array having the same length as the container (see count). The method will delete those elements whose corresponding mask value is non-zero, i.e., the i-th element will be deleted if mask[i]!=0.

For example, to delete all currently selected particles, i.e., the subset of particles whose Selection property is non-zero, one would simply write:

data.particles_.delete_elements(data.particles['Selection'])

The effect of this statement is the same as for applying the DeleteSelectedModifier to the particles list.

delete_indices(indices)

Deletes a subset of the elements from this container. The elements to be deleted must be specified in terms of a sequence of indices, all in the range 0 to count-1. The method accepts any type of iterable object, including sequence types and generators.

For example, to delete every other particle, one could use Python’s range() function to generate all even indices up to the length of the particle container:

data.particles_.delete_indices(range(0, data.particles.count, 2))
property title

The title of the data object under which it appears in the user interface of OVITO.

class ovito.data.SimulationCell

Base: ovito.data.DataObject

This object stores the geometric shape and boundary conditions of the simulation box. Typically there is exactly one SimulationCell object in a DataCollection, which is accessible through the cell field:

data = pipeline.compute()
print(data.cell[...])   # Use [...] to cast SimulationCell object to a NumPy array

The cell matrix

The geometry of the simulation cell is encoded as a 3x4 matrix \(\mathbf{M}\). The first three columns \(\mathbf{a}\), \(\mathbf{b}\), \(\mathbf{c}\) of the matrix are the vectors spanning the three-dimensional parallelepiped in Cartesian space. The fourth column specifies the Cartesian coordinates of the cell’s origin \(\mathbf{o}\) within the global simulation coordinate system:

\[\begin{split}\mathbf{M} = \begin{pmatrix} a_x & b_x & c_x & o_x \\ a_y & b_y & c_y & o_y \\ a_z & b_z & c_z & o_z \\ \end{pmatrix}\end{split}\]

The cell matrix is represented by a two-dimensional NumPy array of shape (3,4) using row-major storage order:

a = data.cell[:,0]
b = data.cell[:,1]
c = data.cell[:,2]
o = data.cell[:,3]

The is2D flag of the simulation cell indicates whether the system is two-dimensional. The cell matrix of a 2d system also has the 3x4 shape, but the cell vector \(\mathbf{c}\) and the last row of the cell matrix are ignored by many computations in OVITO if the system is marked as 2d.

Periodic boundary conditions

The pbc field stores a tuple of three Boolean flags that indicate for each cell vector whether the system is periodic in that direction or not. OVITO uses that information in various computations. If the system is two-dimensional, the value of the third pbc flag is ignored.

Modifying the simulation cell

When you modify the entries of the cell matrix, make sure you use the underscore notation to request a modifiable version of the SimulationCell object:

# Make cell twice as large along the Y direction by scaling the second cell vector:
data.cell_[:,1] *= 2.0

Reset the simulation cell to an orthogonal box \((L_x, L_y, L_z)\) centered at the origin:

lx = 20.0; ly = 10.0; lz = 8.0
data.cell_[:,0] = (lx, 0, 0)
data.cell_[:,1] = (0, ly, 0)
data.cell_[:,2] = (0, 0, lz)
data.cell_[:,3] = numpy.dot((-0.5, -0.5, -0.5), data.cell[:3,:3])
data.cell_.pbc = (True, True, True)

Conversion between Cartesian and reduced coordinates

Given a point in 3d space, \(\mathbf{p}=(x, y, z)\), expressed in coordinates of the Cartesian simulation system, you can compute the corresponding reduced cell coordinates by extending the point to a quadruplet \((x, y, z, 1)\) and multiplying it with the inverse cell matrix \(\mathbf{M}^*\):

p_cartesian = (x, y, z)
p_reduced   = cell.inverse @ numpy.append(p_cartesian, 1.0)  # @-operator is shorthand for numpy.matmul()

This effectively performs an affine transformation. The reverse transformation back to Cartesian coordinates in the global simulation system works in the same way. The following operation converts a 3d point from reduced cell coordinates to simulation coordinates:

p_reduced   = (xs, ys, zs)
p_cartesian = cell @ numpy.append(p_reduced, 1.0)  # @-operator is shorthand for numpy.matmul()

Transforming vectors (as opposed to points) between Cartesian and reduced cell coordinates works somewhat differently, because vectors are not affected by the translation of the simulation cell, i.e., when the cell’s origin does not coincide with the origin of the global simulation coordinate system. A vector \(\mathbf{v}=(x, y, z)\) should thus be amended with a zero, \((x, y, z, 0)\), before applying the 3x4 transformation matrix to ignore the translational component:

v_cartesian     = (vx, vy, vz)
v_reduced       = cell.inverse @ numpy.append(v_cartesian, 0.0)
v_cartesian_out = cell @ numpy.append(v_reduced, 0.0)  
assert numpy.allclose(v_cartesian_out, v_cartesian)

The operations described above transform individual 3d points or vectors. In case you have to transform an entire array of points or vectors, for example the list of atomic positions, it is most efficient to apply the transformation to all elements of the array at once. Here is how you can do the affine transformation back and forth between Cartesian and reduced coordinates for an array:

cartesian_positions     = data.particles.positions
reduced_positions       = (cell.inverse[0:3,0:3] @ cartesian_positions.T).T + cell.inverse[0:3,3]
cartesian_positions_out = (cell[0:3,0:3] @ reduced_positions.T).T + cell[0:3,3]
assert numpy.allclose(cartesian_positions_out, cartesian_positions)

When transforming an array of vectors, leave away the translation term and perform just the linear transformation (3x3 matrix-vector multiplication).

Visual representation

Each SimulationCell object has an attached SimulationCellVis element, which controls the visual appearance of the wireframe box in rendered images. It can be accessed via the vis attribute inherited from the DataObject base class:

data = pipeline.compute()
# Change display color of simulation cell to red:
data.cell.vis.rendering_color = (1.0, 0.0, 0.0)
# Or turn off the display of the cell completely:
data.cell.vis.enabled = False
delta_vector(ra, rb, return_pbcvec=False)

Computes the vector connecting two points \(r_a\) and \(r_b\) in a periodic simulation cell by applying the minimum image convention.

The method starts by computing the 3d vector \({\Delta} = r_b - r_a\) for two points \(r_a\) and \(r_b\), which may be located in different images of the periodic simulation cell. The minimum image convention is then applied to obtain the new vector \({\Delta'} = r_b' - r_a\), where the original point \(r_b\) has been replaced by the periodic image \(r_b'\) that is closest to \(r_a\), making the vector \({\Delta'}\) as short as possible (in reduced coordinate space). \(r_b'\) is obtained by translating \(r_b\) an integer number of times along each of the three cell directions: \(r_b' = r_b - H*n\), with \(H\) being the 3x3 cell matrix and \(n\) being a vector of three integers that are chosen by the method such that \(r_b'\) is as close to \(r_a\) as possible.

Note that the periodic image convention is applied only along those cell directions for which periodic boundary conditions are enabled (see pbc property). For other directions no shifting is performed, i.e., the corresponding components of \(n = (n_x,n_y,n_z)\) will always be zero.

The method is able to compute the results for either an individual pair of input points or for two arrays of input points. In the latter case, i.e. if the input parameters ra and rb are both 2-D arrays of shape Nx3, the method returns a 2-D array containing N output vectors. This allows applying the minimum image convention to a large number of point pairs in one function call.

The option return_pbcvec lets the method return the vector \(n\) introduced above as an additional output. The components of this vector specify the number of times the image point \(r_b'\) needs to be shifted along each of the three cell directions in order to bring it onto the original input point \(r_b\). In other words, it specifies the number of times the computed vector \({\Delta} = r_b - r_a\) crosses a periodic boundary of the cell (either in positive or negative direction). For example, the PBC shift vector \(n = (1,0,-2)\) would indicate that, in order to get from input point \(r_a\) to input point \(r_b\), one has to cross the cell boundaries once in the positive x-direction and twice in the negative z-direction. If return_pbcvec is True, the method returns the tuple (\({\Delta'}\), \(n\)); otherwise it returns just \({\Delta'}\). Note that the vector \(n\) computed by this method can be used, for instance, to correctly initialize the Bonds.pbc_vectors property for newly created bonds that cross a periodic cell boundary.

Parameters:
  • ra – The Cartesian xyz coordinates of the first input point(s). Either a 1-D array of length 3 or a 2-D array of shape (N,3).

  • rb – The Cartesian xyz coordinates of the second input point(s). Must have the same shape as ra.

  • return_pbcvec (bool) – If True, also returns the vector \(n\), which specifies how often the vector \((r_b' - r_a)\) crosses the periodic cell boundaries.

Returns:

The vector \({\Delta'}\) and, optionally, the vector \(n\).

Note that there exists also a convenience method Particles.delta_vector(), which should be used in situations where \(r_a\) and \(r_b\) are the coordinates of two particles in the simulation cell.

property inverse

Read-only property returning the reciprocal cell matrix \(\mathbf{M}^*\) - an array of shape (3,4):

\[\mathbf{M}^* = \begin{bmatrix} \mathbf{a}^* & \mathbf{b}^* & \mathbf{c}^* & \mathbf{o}^* \end{bmatrix}\]

with the real-space cell volume \(V = (\mathbf{a} \times \mathbf{b}) \cdot \mathbf{c}\) and reciprocal cell vectors given by

\[\mathbf{a}^* = \frac{\mathbf{b} \times \mathbf{c}}{V} \qquad \mathbf{b}^* = \frac{\mathbf{c} \times \mathbf{a}}{V} \qquad \mathbf{c}^* = \frac{\mathbf{a} \times \mathbf{b}}{V} \qquad \mathbf{o}^* = -\begin{pmatrix} \mathbf{a}^* \; \mathbf{b}^* \; \mathbf{c}^* \end{pmatrix} \mathbf{o} \mathrm{.}\]
property is2D

Specifies whether the system is two-dimensional (instead of three-dimensional). For two-dimensional systems, the third pbc flag and the cell vector \(\mathbf{c}\) are typically ignored.

Default:

False

property pbc

A tuple of three Boolean flags specifying whether periodic boundary conditions are enabled along the cell’s three spatial directions.

Default:

(False, False, False)

property volume

Read-only property computing the volume of the three-dimensional simulation cell. The returned value is equal to the absolute determinant of the 3x3 submatrix formed by the three cell vectors, i.e. the scalar triple product \(V=|(\mathbf{a} \times \mathbf{b}) \cdot \mathbf{c}|\):

assert cell.volume == abs(numpy.linalg.det(cell[0:3,0:3]))
property volume2D

Read-only property computing the area of the two-dimensional simulation cell (see is2D). The returned value is equal to the magnitude of the cross-product of the first two cell vectors, i.e. \(V_{\mathrm{2d}} = |\mathbf{a} \times \mathbf{b}|\):

assert cell.volume2D == numpy.linalg.norm(numpy.cross(cell[:,0], cell[:,1]))
class ovito.data.SurfaceMesh

Base: ovito.data.DataObject

This data object type represents a surface in three-dimensional space, i.e.. a two-dimensional manifold that is usually closed and orientable. The underlying representation of the surface is a discrete mesh made of vertices, edges, and planar faces. See the user manual page on surface meshes for more information on this data object type.

Surface meshes are typically produced by modifiers such as ConstructSurfaceModifier, CreateIsosurfaceModifier, CoordinationPolyhedraModifier or VoronoiAnalysisModifier.

Each surface mesh has a unique identifier by which it can be looked up in the DataCollection.surfaces dictionary:

# Apply a CreateIsosurfaceModifier to a VoxelGrid to create a SurfaceMesh:
pipeline.modifiers.append(CreateIsosurfaceModifier(operate_on='voxels:charge-density', property='Charge density', isolevel=0.05))
data = pipeline.compute()

# The SurfaceMesh created by the modifier has the identifier 'isosurface':
surface = data.surfaces['isosurface']

Vertices, halfedges, and faces

A surface mesh is made of a set of vertices, a set of directed halfedges each connecting two vertices, and a set of faces, each formed by a circular sequence of halfedges. The connectivity information, i.e., which vertices are connected by halfedges and which halfedges form the faces, is stored in the topology sub-object of the SurfaceMesh. See the SurfaceMeshTopology class for more information.

Vertices and faces of the surface mesh can be associated with arbitrary property values, similar to how particles can have arbitrary properties assigned to them in OVITO. These properties are managed by the vertices and faces PropertyContainer sub-objects of the surface mesh. The vertices of the mesh are always associated with the property named Position, which stores the three-dimensional coordinates of each vertex, similar to the Position property of particles in OVITO.

vertex_coords = surface.vertices['Position']

The SurfaceMeshVis element, which is responsible for rendering the surface mesh, provides the option to visualize local vertex and face property values using a color mapping scheme.

surface.vis.color_mapping_mode = SurfaceMeshVis.ColorMappingMode.Vertex
surface.vis.color_mapping_property = 'Position.Z'
surface.vis.color_mapping_interval = (min(vertex_coords[:,2]), max(vertex_coords[:,2]))

If you want to modify property values of the mesh, keep in mind that you have to use underscore notation, for example:

data.surfaces['isosurface_'].vertices_['Position_'] += (xoffset, yoffset, zoffset)

Periodic simulation domains

A surface mesh may be embedded in a periodic domain, i.e. in a simulation cell with periodic boundary conditions. That means edges and faces of the surface mesh can connect vertices on opposite sides of the simulation box and will wrap around correctly. OVITO takes care of computing the intersections of such a periodic surface with the box boundaries and automatically produces a non-periodic representation of the mesh when it comes to displaying the surface. If needed, you can explicitly request a non-periodic version of the mesh, which was clipped at the periodic box boundaries, by calling the to_triangle_mesh() method from a script.

The spatial domain of the surface mesh is the SimulationCell object stored in its SurfaceMesh.domain field. Note that this attached SimulationCell may, in some situations, not be identical with the global simulation cell set for the DataCollection.

Spatial regions

If it is a closed, orientable manifold the surface mesh subdivides three-dimensional space into separate spatial regions. For example, if the surface mesh was constructed by the ConstructSurfaceModifier from a set of input particles, then the volume enclosed by the surface is the “filled” interior region and the exterior space is the “empty” region containing no particles.

In general, the SurfaceMesh class manages a variable list of regions, each being identified by a numeric, zero-based index. The locate_point() method allows to determine which spatial region some point in space belongs to.

A surface mesh may be degenerate, which means it contains no vertices and faces. In such a case there is only one spatial region filling entire space. For example, when there exist no input particles, the ConstructSurfaceModifier is unable to construct a regular surface mesh and the “empty” region fills the entire simulation cell. Conversely, if the periodic simulation cell is completely filled with particles, the “filled” region covers the entire periodic simulation domain and the resulting surface mesh consists of no vertices or faces, i.e., it is also degenerate. To discriminate between the two situations, the SurfaceMesh class has a space_filling_region field, which specifies the spatial region that fills entire space in cases where the mesh is degenerate.

File export

A surface mesh may be exported to a geometry file in the form of a triangle mesh using OVITO’s export_file() function. To this end, a non-periodic version is produced by truncating triangles at the domain boundaries and generating “cap polygons” filling the holes that occur at the intersection of the surface with periodic domain boundaries. The following example code writes a VTK geometry file (vtk/trimesh export format):

from ovito.io import import_file, export_file
from ovito.data import SurfaceMesh
from ovito.modifiers import ConstructSurfaceModifier

# Load a particle set and construct the surface mesh:
pipeline = import_file("input/simulation.dump")
pipeline.modifiers.append(ConstructSurfaceModifier(radius = 2.8))
mesh = pipeline.compute().surfaces['surface']

# Export the mesh to a VTK file for visualization with ParaView.
export_file(mesh, 'output/surface_mesh.vtk', 'vtk/trimesh')

Clipping planes

A set of clipping planes can be assigned to a SurfaceMesh to clip away parts of the mesh for visualization purposes. This may be useful to e.g. cut a hole into a closed surface allowing to look inside the enclosed volume. The SurfaceMesh objects manages a list of clipping planes, which is accessible through the get_clipping_planes() and set_clipping_planes() methods. Note that the cut operations are non-destructive and get performed only on the transient, non-periodic version of the mesh generated during image rendering or when exporting the mesh to a file. The original surface mesh is not affected. The SliceModifier, when applied to a SurfaceMesh, performs the slice by simply adding a corresponding clipping plane to the SurfaceMesh. The actual truncation of the mesh happens later on, during the final visualization step, when a non-periodic version is computed.

connect_opposite_halfedges() bool

Links together pairs of halfedges in the mesh to form a two-dimensional manifold made of connected faces. For each halfedge \(a \to b\) the method tries to find the corresponding reverse halfedge \(b \to a\), which bounds the adjacent face. The two halfedges are then linked together to form a pair. The method returns True to indicate that all halfedges of the mesh have been successfully associated with a corresponding opposite halfedge. In this case, the mesh is said to be closed, i.e., its faces form a contiguous manifold.

Important

For this method to work, the faces of the mesh must have been created all with the same winding order. That means their vertices must consistently be ordered counter-clockwise when viewed from the outside of the closed surface manifold (the front side). Only then do the halfedges of adjacent faces run in opposite directions and can be successfully paired by this method.

New in version 3.7.9.

create_face(vertices: Sequence[int] | numpy.ndarray) int

Adds a new face to the mesh. vertices must be a sequence of two or more zero-based indices into the mesh’s vertex list. The method creates a loop of halfedges connecting the given vertices to form a closed polygon. The zero-based index of the newly created face is returned.

Tip

If you intend to add several faces to the mesh, consider using create_faces() instead. It is potentially much faster than calling create_face() multiple times.

Note

Visible faces should be made of three or more vertices that form a convex polygon. Faces that represent a non-convex polygon will likely be rendered incorrectly by OVITO. Faces having only two edges, while technically valid, will not get rendered because they are degenerate.

The vertex winding order used by OVITO is counter-clockwise on the front side of mesh faces. When constructing a closed mesh, make sure you always specify vertices in counter-clockwise order when viewed from the outside of enclosed region.

Code example:

# Add a new SurfaceMesh object to the DataCollection with unique object identifier 'quad'.
# The simulation cell of the particle system is adopted also as domain of the SurfaceMesh.
mesh = data.surfaces.create(identifier='quad', title='Quad', domain=data.cell)

# Create 4 mesh vertices forming a quadrilateral.
verts = [[0,0,0], [10,0,0], [10,10,0], [0,10,0]]
mesh.create_vertices(verts)

# Create a face connecting the 4 vertices.
mesh.create_face([0,1,2,3])

# Initialize the 'Color' property of the newly created face.
mesh.faces.create_property('Color', data=[(1,0,0)])

New in version 3.7.9.

create_faces(vertex_lists: Sequence[Sequence[int]] | numpy.ndarray) int

Adds several new polygonal faces to the mesh.

Parameters:

vertex_lists – A sequence of sequences, one for each face to be created, which specify the vertex indices to be connected by the new mesh faces.

Returns:

Index of the first newly created face.

vertex_lists may be list of tuples for example. The following call creates a 3-sided and a 4-sided polygonal face:

mesh.create_faces([(0,1,2), (3,4,5,6)])

For best performance, pass a two-dimensional NumPy array to create multiple faces which all have the same number of vertices:

# Nx3 array [[0,1,2], [3,4,5], [6,7,8], ...] for connecting 3N vertices with triangle faces.
triangle_list = numpy.arange(mesh.vertices.count).reshape((mesh.vertices.count//3, 3))
mesh.create_faces(triangle_list)

A third option is to specify the faces as one linear array, in which each face’s vertex list is prefixed with the number of vertices. For example, to create a 3-sided face 0-1-2 and a 4-sided face 3-4-5-6, one would write:

mesh.create_faces(numpy.asarray([3,0,1,2,  4,3,4,5,6]))

Note that the data must be provided as a NumPy array in this case, not a Python list.

The create_faces() method has two effects: It increments the mesh’s topology.face_count and it extends the arrays in the mesh’s faces property container, which stores all per-face properties. The method raises an error if any of the specified vertex indices does not exist in the mesh. That means you should first call create_vertices() to add vertices to the mesh before creating faces referencing these vertices.

Note

Visible faces should be made of three or more vertices forming convex polygons. Faces that represent non-convex polygons will likely be rendered incorrectly by OVITO. Faces having only two edges, while technically valid, will not get rendered because they are degenerate.

The vertex winding order used by OVITO is counter-clockwise on the front side of mesh faces. When constructing a closed mesh, make sure you always specify vertices in counter-clockwise order when viewed from the outside of enclosed region.

Usage example:

# Add a new SurfaceMesh object to the DataCollection with unique object identifier 'tetrahedron'.
# The simulation cell of the particle system is adopted also as domain of the SurfaceMesh.
mesh = data.surfaces.create(identifier='tetrahedron', title='Tetrahedron', domain=data.cell)

# Create 4 mesh vertices.
verts = [[0,0,0], [10,0,0], [0,10,0], [0,0,10]]
mesh.create_vertices(verts)

# Create 4 triangular faces forming a tetrahedron.
mesh.create_faces([[0,1,2], [0,2,3], [0,3,1], [1,3,2]])

# Initialize the 'Color' property of the newly created faces with RGB values.
mesh.faces.create_property('Color', data=[(1,0,0), (1,1,0), (0,0,1), (0,1,0)])

# Make it a "closed" mesh, connecting the four faces to form a surface manifold.
mesh.connect_opposite_halfedges()

New in version 3.7.9.

create_vertices(coords: Sequence[Sequence[float]] | numpy.ndarray) int

Adds a set of new vertices to the mesh. coords must be an \(n \times 3\) array specifying the xyz coordinates of the \(n\) vertices to create. The coordinates will be copied into the Position vertex property, which is managed by the vertices property container. Furthermore, the vertex_count value of the mesh’s topology will be incremented by \(n\).

Initially, the new vertices will not be associated with any faces. Use create_face() or create_faces() to create faces connecting the vertices.

New in version 3.7.9.

property domain

The SimulationCell describing the (possibly periodic) domain which this surface mesh is embedded in. Note that this cell generally is independent of and may be different from the cell found in the DataCollection.

property faces

The PropertyContainer storing the per-face properties of the mesh.

In general, an arbitrary set of uniquely named properties may be associated with the faces of a surface mesh. OVITO defines the following standard face properties, which have a well-defined meaning and prescribed data layout:

Standard property name

Data type

Component names

Color

float32

R, G, B

Region

int32

Selection

int8

The property Color can be set to give each face of the surface mesh an individual color. It overrides the uniform coloring otherwise controlled by the SurfaceMeshVis element.

The property Region links each face with the volumetric region of the SurfaceMesh that it bounds (see description above). The values of this property are zero-based indices into the regions list of the mesh.

The property Selection controls the selection state of each individual mesh face. This property is set by modifiers that create selections, such as ExpressionSelectionModifier, and is used by modifiers that operate on the subset of currently selected faces, such as AssignColorModifier. All faces whose Selection property has a non-zero value are part of the current selection set.

get_clipping_planes()

Returns an \(N \times 4\) array containing the definitions of the N clipping planes attached to this SurfaceMesh.

Each plane is defined by its unit normal vector and a signed displacement magnitude, which determines the plane’s distance from the coordinate origin along the normal, giving four numbers per plane in total. Those parts of the surface mesh which are on the positive side of the plane (in the direction the normal vector) will be cut away during rendering.

Note that the returned Numpy array is a copy of the internal data stored by the SurfaceMesh.

get_face_vertices(flat=False) numpy.ndarray | list[list[int]]

Returns an array with the vertex indices of all mesh faces. The parameter flat controls how the face vertices get returned by the function:

flat=False: If all \(n\) faces of the mesh have the same, uniform number of vertices, \(m\), for example, if they are all triangles, then a 2-d NumPy array of shape \((n, m)\) containing the zero-based vertex indices is returned. Otherwise, a list of lists is returned, in which nested lists may have different lengths.

flat=True: Returns a 1-d array with the vertex lists of all mesh faces stored back to back. A face’s vertex list is preceded by the number of vertices of that face. Then the actual vertex indices of the face follow. Then the number of vertices of the next face follows, and so on.

New in version 3.7.9.

locate_point(pos, eps=1e-6)

Determines which spatial region of the mesh contains the given point in 3-d space.

The function returns the numeric ID of the region pos is located in. Note that region ID -1 is typically reserved for the empty exterior region, which, if it exists, is the one not containing any atoms or particles. Whether non-negative indices refer to only filled (interior) regions or also empty regions depends on the algorithm that created the surface mesh and its spatial regions.

The parameter eps is a numerical precision threshold to detect if the query point is positioned exactly on the surface boundary, i.e. on the manifold separating two spatial regions. This condition is indicated by the special return value None. Set eps to 0.0 to disable the point-on-boundary test. Then the method will never return None as a result, but the determination of the spatial region will become numerically unstable if the query point is positioned right on a boundary surface.

Parameters:
  • pos (array-like) – The (x,y,z) coordinates of the query point

  • eps (float) – Numerical precision threshold for point-on-boundary test

Returns:

The numeric ID of the spatial region containing pos; or None if pos is exactly on the dividing boundary between two regions

property regions

The PropertyContainer storing the properties of the spatial regions of the mesh.

In general, an arbitrary set of uniquely named properties may be associated with the regions of a surface mesh. OVITO defines the following standard region properties, which have a well-defined meaning and prescribed data layout:

Standard property name

Data type

Component names

Color

float32

R, G, B

Filled

int8

Selection

int8

Surface Area

float64

Volume

float64

The property Color can be set to give the faces bounding each of the volumetric regions a different color. It overrides the uniform mesh coloring otherwise controlled by the SurfaceMeshVis element.

The property Filled is a flag indicating for each region whether it is an interior region, e.g. inside a solid, or an empty exterior region, e.g. outside the solid bounded by the surface. This property is created by the ConstructSurfaceModifier. The same is true for the per-region properties Surface Area and Volume.

set_clipping_planes(planes)

Sets the clipping planes of this SurfaceMesh. The array planes must follow the same format as the one returned by get_clipping_planes().

property space_filling_region

Indicates the index of the spatial region that fills the entire domain in case the surface is degenerate, i.e. the mesh has zero faces. The invalid index -1 is typically associated with the empty (exterior) region.

to_triangle_mesh() Tuple[TriangleMesh, TriangleMesh, numpy.ndarray]

Converts the surface into a non-periodic TriangleMesh.

Returns:

(trimesh, caps, facemap)

  • trimesh: A TriangleMesh representing the surface geometry after clipping it at the periodic boundaries of the domain and any attached clipping planes (see get_clipping_planes()).

  • caps: A TriangleMesh containing the cap polygons generated at intersections of the periodic surface mesh with boundaries of the simulation domain. Will be None if the surface mesh has no attached domain, the domain is degenerate, or the surface mesh does not represent a closed manifold.

  • facemap: A NumPy array of indices into the face list of this SurfaceMesh, one for each triangular face of the TriangleMesh trimesh. This map lets you look up for each face of the output mesh what the corresponding face of the input surface mesh is.

New in version 3.7.5.

property topology

A SurfaceMeshTopology object storing the face connectivity of the mesh.

property vertices

The PropertyContainer storing all per-vertex properties of the mesh, including the vertex coordinates.

In general, an arbitrary set of uniquely named properties may be associated with the vertices of a surface mesh. OVITO defines the following standard vertex properties, which have a well-defined meaning and prescribed data layout:

Standard property name

Data type

Component names

Color

float32

R, G, B

Position

float64

X, Y, Z

Selection

int8

The property Position is always present and stores the Cartesian vertex coordinates.

The property Color can be set to give each vertex of the surface mesh an individual color. It overrides the uniform coloring otherwise controlled by the SurfaceMeshVis element. Vertex colors get interpolated across the mesh faces during rendering.

The property Selection controls the selection state of each individual mesh vertex. This property is set by modifiers that create selections, such as ExpressionSelectionModifier, and is used by modifiers that operate on the subset of currently selected vertices, such as AssignColorModifier. All vertices whose Selection property has a non-zero value are part of the current selection set.

class ovito.data.SurfaceMeshTopology

Base: ovito.data.DataObject

New in version 3.7.6.

This data structure holds the connectivity information of a SurfaceMesh. It is accessible through the SurfaceMesh.topology field. The surface mesh topology consists of vertices, faces and halfedges.

All these topological entities of the mesh are identified by numeric indices ranging from 0 to (vertex_count-1), (face_count-1), and (edge_count-1), respectively. The vertices and faces of the mesh may be associated with auxiliary properties, which are stored separately from the topology in the SurfaceMesh.vertices and SurfaceMesh.faces property containers. In particular, the spatial coordinates of the mesh vertices are stored as Position property array in SurfaceMesh.vertices.

../_images/halfedge_structure.png

A halfedge is a directed edge \(a \to b\) connecting two vertices \(a\) and \(b\) – depicted as a half-arrow in the figure. A face is implicitly defined by a circular sequence of halfedges that bound the face. Typically, halfedges come in pairs. The halfedge \(a \to b\) and its opposite halfedge, \(b \to a\), form a pair that links two neighboring faces together. Thus, halfedge pairs are essential for forming a connected, two-dimensional surface manifold. The surface is said to be closed, i.e., it has no open boundaries if all halfedges of the mesh are associated with corresponding opposite halfedges (see is_closed).

For each vertex the topology object maintains a linked-list of directed halfedges leaving that vertex. It can be accessed through the first_vertex_edge() and next_vertex_edge() methods.

For each face the topology object maintains a circular linked-list of directed halfedges bounding that face (in counter-clockwise winding order). It can be accessed through the first_face_edge() and next_face_edge()/prev_face_edge() methods.

Tip

All query methods of this class are vectorized, which means they are able to process an array of arguments in a single function call and will return a corresponding array of results. The advantage of this is that the loop over the elements in the argument array runs entirely on the C++ side, which is typically much faster than a for-loop in Python. For example, to generate a list with the first halfedge of every mesh face:

# Version 1: vectorized function call (fast)
edges = mesh.topology.first_face_edge(range(mesh.topology.face_count))

# Version 2: explicit loop (slow)
edges = [mesh.topology.first_face_edge(face) for face in range(mesh.topology.face_count)]
adjacent_face(edge: int) int

Returns the face that is bounded by the halfedge edge.

count_face_edges(face: int) int

Returns the number of halfedges that bound face. See the code example for next_face_edge() to learn how this method works.

count_manifolds(edge: int) int
count_vertex_edges(vertex: int) int

Returns the number of halfedges that leave vertex. See the code example for next_vertex_edge() to learn how this method works.

property edge_count

Total number of halfedges in the SurfaceMesh. This property is read-only. Halfedges are created automatically by SurfaceMesh.create_face() or SurfaceMesh.create_faces() when adding new faces to the mesh topology.

property face_count

Number of faces in the SurfaceMesh. This is always equal to the count of the SurfaceMesh.faces property container.

This property is read-only. Use SurfaceMesh.create_face() or SurfaceMesh.create_faces() to add new faces to the mesh.

find_edge(face: int, vertex1: int, vertex2: int) int

Given a face, finds the halfedge of that face which leads from vertex1 to vertex2. If no such halfedge exists, returns -1.

This method can be used to quickly find the edge connecting two vertices of a face without the need to explicitly visit and check each edge bounding the face.

first_edge_vertex(edge: int) int

Returns the vertex the given halfedge is leaving from. To retrieve the vertex the halfedge is leading to, call second_edge_vertex().

first_face_edge(face: int) int

Returns some halfedge bounding the given face. Given that first halfedge, all other halfedges bounding the same face can be visited using next_face_edge() or prev_face_edge().

first_face_vertex(face: int) int

Given a face, this method returns some vertex of that face. This is equivalent to retrieving the vertex to which the first halfedge of the face is connected to, i.e. first_edge_vertex(first_face_edge(face)).

first_vertex_edge(vertex: int) int

Returns the head halfedge from the linked list of halfedges leaving vertex. Subsequent halfedges from the linked list can be retrieved with next_vertex_edge(). If no halfedges are connected to vertex, the method returns -1.

has_opposite_edge(edge: int) bool

Returns whether the given halfedge edge is associated with a corresponding reverse halfedge bounding an adjacent face in the same manifold. This is equivalent to checking the return value of opposite_edge(), which returns -1 to indicate that edge does not have an opposite edge.

has_opposite_face(face: int) bool

Returns whether face is part of a two-sided manifold. A face that is part of a two-sided manifold has a ‘partner’ face with opposite orientation, which can be retrieved through the opposite_face() method.

property is_closed

This is a read-only property indicating whether the surface mesh is fully closed. In a closed mesh, all faces are connected to exactly one adjacent face along each of their halfedges. That means the mesh presents a two-dimensional surface manifold without borders. Furthermore, a closed mesh divides space into an “interior” and an “exterior” region.

New in version 3.7.9.

next_face_edge(edge: int) int

Given the halfedge edge bounding some face, this method returns the following halfedge when going around the face in forward direction (counter-clockwise - when looking at the front side of the face). All halfedges of the face form a circular sequence - without a particular beginning or end. You can loop over this circular sequence in forward or backward direction with the next_face_edge() and prev_face_edge() methods. Given some mesh face, you can obtain a first halfedge through the first_face_edge() method.

The following code example shows how to visit all halfedges of a face in order. Since the halfedges form a circular linked list, we have to remember which edge we started from to terminate the loop once we reach the first edge again:

def count_edges(mesh: SurfaceMesh, face: int) -> int:
    start_edge = mesh.topology.first_face_edge(face)
    count = 1
    edge = mesh.topology.next_face_edge(start_edge)
    while edge != start_edge:
        assert mesh.topology.adjacent_face(edge) == face
        count += 1
        edge = mesh.topology.next_face_edge(edge)
    return count

# The function defined above is equivalent to SurfaceMeshTopology.count_face_edges():
assert count_edges(mesh, 0) == mesh.topology.count_face_edges(0)
next_manifold_edge(edge: int) int
next_vertex_edge(edge: int) int

Returns another halfedge leaving from the same vertex as edge. Together with first_vertex_edge() this method allows you to iterate over all halfedges connected to some vertex. When the end of the vertex’ edge list has been reached, the method returns -1.

The following example demonstrates how to visit all outgoing halfedges of some vertex and count them:

def count_edges(mesh: SurfaceMesh, vertex: int) -> int:
    count = 0
    edge = mesh.topology.first_vertex_edge(vertex)
    while edge != -1:
        assert mesh.topology.first_edge_vertex(edge) == vertex
        count += 1
        edge = mesh.topology.next_vertex_edge(edge)
    return count

# The function defined above is equivalent to SurfaceMeshTopology.count_vertex_edges():
assert count_edges(mesh, 0) == mesh.topology.count_vertex_edges(0)
opposite_edge(edge: int) int

Given the halfedge edge, returns the reverse halfedge that bounds an adjacent face. This opposite halfedge connects the same two vertices as edge but in reverse order. You can use this method to determine whether the face bounded by edge has a neighboring face that is part of the same manifold:

def get_neighboring_face(mesh: SurfaceMesh, edge: int) -> int:
    opp_edge = mesh.topology.opposite_edge(edge)
    if opp_edge == -1: return -1
    assert mesh.topology.first_edge_vertex(edge) == mesh.topology.second_edge_vertex(opp_edge)
    assert mesh.topology.second_edge_vertex(edge) == mesh.topology.first_edge_vertex(opp_edge)
    return mesh.topology.adjacent_face(opp_edge)

You may call the convenience method has_opposite_edge() to determine whether a halfedge is associated with a corresponding opposite halfedge. If the surface mesh does not form a closed manifold, the halfedges at the boundary of the manifold do not have opposite halfedges, because there are no adjacent faces where the surface terminates.

opposite_face(face: int) int

Returns the face on the opposite side of the two-sided manifold, or -1 if the manifold is one-sided. The returned face shares the same vertices with face but in reverse order. Note that opposite_face(opposite_face(face))==face.

prev_face_edge(edge: int) int

Given the halfedge edge bounding some face, this method returns the previous halfedge going around that face in backward direction (clockwise - when looking at the front side of the face). All halfedges of a face form a circular sequence - without a particular beginning or end. You can loop over this circular sequence in forward or backward direction with the next_face_edge() and prev_face_edge() methods.

second_edge_vertex(edge: int) int

Returns the vertex the given halfedge is leading to. To retrieve the vertex the halfedge is leaving from, call first_edge_vertex().

property vertex_count

Number of vertices in the SurfaceMesh. This is always equal to the count of the SurfaceMesh.vertices property container.

This property is read-only. Use SurfaceMesh.create_vertices() to add new vertices to the mesh.

class ovito.data.TriangleMesh

Base: ovito.data.DataObject

This data object type stores a three-dimensional mesh made of vertices and triangular faces. Such a mesh can describing general polyhedral geometry to be visualized side by side with the particle simulation data.

Typically, triangle meshes are imported from external geometry data files (e.g. STL, OBJ, VTK formats) using the import_file() function. See also the corresponding section of the OVITO user manual. All triangle meshes in a data collection are accessible through the DataCollection.triangle_meshes dictionary view.

Note that the SurfaceMesh class is a second object type that can represent surface geometries, just like a TriangleMesh. In contrast to triangle meshes, surface meshes may be embedded in periodic simulation domains and are closed manifolds in most cases. Furthermore, surface meshes can store arbitrary per-vertex and per-face property values – something triangle meshes cannot do. A triangle mesh is a more low-level data structure, which can be sent directly to a GPU for rendering. A surface mesh, in contrast, is a more high-level data structure, which gets automatically converted to a triangle mesh for visualization.

The visual appearance of the triangle mesh is controlled through the attached TriangleMeshVis element (see DataObject.vis field of base class).

A triangle mesh consists of \(n_{\mathrm{v}}\) vertices and \(n_{\mathrm{f}}\) triangular faces. These counts are exposed by the class as attributes vertex_count and face_count. Each face connects three vertices of the mesh, and several faces may share a vertex. The faces are stored as triplets of zero-based indices into the vertex list.

property face_count

The number of triangular faces of the mesh, \(n_{\mathrm{f}}\).

Default:

0

get_faces()

Returns the list of triangles of the mesh as a NumPy array of shape \((n_{\mathrm{f}}, 3)\). The array contains for each face three zero-based indices into the mesh’s vertex list as returned by get_vertices(). The returned face array holds a copy of the internal data managed by the TriangleMesh.

get_vertices()

Returns the xyz coordinates of the vertices of the mesh as a NumPy array of shape \((n_{\mathrm{v}}, 3)\). The returned array holds a copy of the internal vertex data managed by the TriangleMesh.

set_faces(vertex_indices)

Sets the faces of the mesh. vertex_indices must be an array-like object of shape \((n_{\mathrm{f}}, 3)\) containing one integer triplet per triangular face. Each integer is a zero-based index into the mesh’s vertex list. The TriangleMesh copies the data from the array into its internal storage. If necessary, the value of face_count is automatically adjusted to match the first dimension of the vertex_indices array.

set_vertices(coordinates)

Sets the xyz coordinates of the vertices of the mesh. coordinates must be an array-like object of shape \((n_{\mathrm{v}}, 3)\). The TriangleMesh copies the data from the array into its internal storage. If necessary, the value of vertex_count is automatically adjusted to match the first dimension of the coordinates array.

property vertex_count

The number of vertices of the mesh, \(n_{\mathrm{v}}\).

Default:

0

class ovito.data.VoxelGrid

Base: ovito.data.PropertyContainer

../_images/grid_example_2d.png

Cell-data grid

A two- or three-dimensional structured grid. Each cell (voxel) of the uniform grid is of the same size and shape. The overall geometry of the grid, its domain, is specified by the attached SimulationCell object, which describes a three-dimensional parallelepiped or a two-dimensional parallelogram. See also the corresponding user manual page for more information on this object type.

The shape property of the grid specifies the number of data points uniformily distributed along each cell vector of the domain. The size of individual voxels depends on the overall domain size as well as the number of data points in each spatial direction. See the grid_type property, which controls whether the data values of the uniform grid are associated with the voxel interiors or the vertices (grid line intersections).

../_images/grid_example_2d_pointdata.png

Point-data grid

Each data point or voxel of the grid may be associated with one or more field values. The data of these voxel properties is stored in standard Property array objects, similar to particle or bond properties. Voxel properties can be accessed by name through the dictionary interface that the VoxelGrid class inherits from its PropertyContainer base class.

Voxel grids can be loaded from input data files, e.g. a CHGCAR file containing the electron density computed by the VASP code, or they can be dynamically generated within OVITO. The SpatialBinningModifier lets you project the information associated with the unstructured particle set to a structured voxel grid.

Given a voxel grid, the CreateIsosurfaceModifier can then generate a SurfaceMesh representing an isosurface for a field quantity defined on the voxel grid.

Example

The following code example demonstrates how to create a new VoxelGrid from scratch and initialize it with data from a NumPy array:

# Starting with an empty DataCollection:
data = DataCollection()

# Create a new SimulationCell object defining the outer spatial dimensions
# of the grid and the boundary conditions, and add it to the DataCollection:
cell = data.create_cell(
    matrix=[[10,0,0,0],[0,10,0,0],[0,0,10,0]],
    pbc=(True, True, True)
)

# Generate a three-dimensional Numpy array containing the grid cell values.
nx = 10; ny = 6; nz = 8
field_data = numpy.random.random((nx, ny, nz))

# Create the VoxelGrid object and give it a unique identifier by which it can be referred to later on.
# Link the voxel grid to the SimulationCell object created above, which defines its spatial extensions.
# Specify the shape of the grid, i.e. the number of cells in each spatial direction.
# Finally, assign a VoxelGridVis visual element to the data object to make the grid visible in the scene.
grid = data.grids.create(
    identifier="field",
    domain=cell,
    shape=(nx,ny,nz),
    grid_type=VoxelGrid.GridType.CellData,
    vis=VoxelGridVis(enabled=True, transparency=0.6)
)

# Add a new property to the voxel grid cells and initialize it with the data from the NumPy array.
# Note that the data must be provided as linear (1-dim.) array with the following type of memory layout:
# The first grid dimension (x) is the fasted changing index while the third grid dimension (z) is the
# slowest varying index. In this example, this corresponds to the "Fortran" memory layout of Numpy.
grid.create_property('Field Value', data=field_data.flatten(order='F'))

# Instead of the flatten() method above, we could also make use of the method VoxelGrid.view()
# to obtain a 3-dimensional view of the property array, which supports direct assignment of grid values.
field_prop = grid.create_property('Field Value', dtype=field_data.dtype, components=1)
grid.view(field_prop)[...] = field_data

# For demonstration purposes, compute an isosurface on the basis of the VoxelGrid created above.
data.apply(CreateIsosurfaceModifier(operate_on='voxels:field', property='Field Value', isolevel=0.7))
property domain

The SimulationCell describing the (possibly periodic) domain which this grid is embedded in. Note that this cell generally is independent of and may be different from the cell found in the DataCollection.

Default:

None

property grid_type

This attribute specifies whether the values stored by the grid object are associated with the voxel cell centers or the grid points (vertices). Possible values are:

  • VoxelGrid.GridType.CellData (default)

  • VoxelGrid.GridType.PointData

A CellData grid represents a field where the sampling points are located in the centers of the voxel cells. This grid type is typically used for volumetric datasets, which represent some quantity within the discrete voxel cell volumes.

../_images/grid_type_celldata.png

A PointData grid represents a field where the sampling points are located at the intersections of the grid lines. Note that, for this grid type only, the type of boundary conditions of the grid’s domain affect the uniform spacing of the sampling points:

../_images/grid_type_pointdata.png

4 x 4 point-data grid (left: non-periodic domain, right: periodic domain)

Default:

VoxelGrid.GridType.CellData

property shape

A 3-tuple specifying the number of sampling points along each of the three cell vectors of the domain.

For two-dimensional grids, for which the is2D property of the domain is set, the third entry in this shape tuple must be equal to 1.

Assigning a new shape to the grid automatically resizes the one-dimensional data arrays stored by this PropertyContainer and updates its PropertyContainer.count property match the product of the three dimensions, i.e. the total number of data points.

Default:

(0, 0, 0)

view(key)

Returns a shaped view of the given grid property, which reflects the 2- or 3-dimensional shape of the grid.

Parameters:

key (str|Property) – The name of the grid property to look up. May include the underscore suffix to make the property mutable. Alternatively, you can directly specify a Property object from this VoxelGrid.

Returns:

A NumPy view of the underlying property array.

Because the VoxelGrid class internally uses linear Property arrays to store the voxel cell values, you normally would have to convert back and forth between the linear index space of the underlying property storage and the 2- or 3-dimensional grid space to access individual voxel cells.

The view() helper method frees you from having to map grid coordinates to array indices because it gives you a shaped NumPy view of the underlying linear storage, which reflects the correct multi-dimensional shape of the grid. For 3-dimensional grids, the ordering of the view’s dimensions is \(x,y,z[,k]\), with \(k\) being an extra dimension that is only present if the accessed property is a vector field quantity. For 2-dimensional grids, the ordering of the view’s dimensions is \(x,y[,k]\).

The returned view lets you conveniently access the values of individual grid cells based on multi-dimensional grid coordinates. Here, as an example, the scalar field property c_ave of a 3-dimensional voxel grid:

nx, ny, nz = grid.shape
field = grid.view('c_ave')
for x in range(nx):
    for y in range(ny):
        for z in range(nz):
            print(field[x,y,z])

New in version 3.9.0.