13.2.7. Mathematical helper functions — MDAnalysis.lib.mdamath
Helper functions for common mathematical operations. Some of these functions are written in C/cython for higher performance.
13.2.7.1. Linear algebra
- MDAnalysis.lib.mdamath.normal(vec1, vec2)[source]
Returns the unit vector normal to two vectors.
\[\hat{\mathbf{n}} = \frac{\mathbf{v}_1 \times \mathbf{v}_2}{|\mathbf{v}_1 \times \mathbf{v}_2|}\]If the two vectors are collinear, the vector \(\mathbf{0}\) is returned.
Changed in version 0.11.0: Moved into lib.mdamath
- MDAnalysis.lib.mdamath.norm(v)[source]
Calculate the norm of a vector v.
\[v = \sqrt{\mathbf{v}\cdot\mathbf{v}}\]This version is faster then numpy.linalg.norm because it only works for a single vector and therefore can skip a lot of the additional fuss linalg.norm does.
- Parameters
v (array_like) – 1D array of shape (N) for a vector of length N
- Returns
norm of the vector
- Return type
- MDAnalysis.lib.mdamath.pdot(a, b)[source]
Pairwise dot product.
a
must be the same shape asb
.- Parameters
a (
numpy.ndarray
of shape (N, M)) –b (
numpy.ndarray
of shape (N, M)) –
- Return type
numpy.ndarray
of shape (N,)
- MDAnalysis.lib.mdamath.pnorm(a)[source]
Euclidean norm of each vector in a matrix
- Parameters
a (
numpy.ndarray
of shape (N, M)) –- Return type
numpy.ndarray
of shape (N,)
- MDAnalysis.lib.mdamath.angle(a, b)[source]
Returns the angle between two vectors in radians
Changed in version 0.11.0: Moved into lib.mdamath
Changed in version 1.0.0: Changed rounding-off code to use np.clip(). Values lower than -1.0 now return np.pi instead of -np.pi
- MDAnalysis.lib.mdamath.dihedral(ab, bc, cd)[source]
Returns the dihedral angle in radians between vectors connecting A,B,C,D.
The dihedral measures the rotation around bc:
ab A---->B \ bc _\' C---->D cd
The dihedral angle is restricted to the range -π <= x <= π.
New in version 0.8.
Changed in version 0.11.0: Moved into lib.mdamath
- MDAnalysis.lib.mdamath.stp(vec1, vec2, vec3)[source]
Takes the scalar triple product of three vectors.
Returns the volume V of the parallel epiped spanned by the three vectors
\[V = \mathbf{v}_3 \cdot (\mathbf{v}_1 \times \mathbf{v}_2)\]Changed in version 0.11.0: Moved into lib.mdamath
- MDAnalysis.lib.mdamath.sarrus_det(matrix)[source]
Computes the determinant of a 3x3 matrix according to the rule of Sarrus.
If an array of 3x3 matrices is supplied, determinants are computed per matrix and returned as an appropriately shaped numpy array.
- Parameters
matrix (numpy.ndarray) – An array of shape
(..., 3, 3)
with the 3x3 matrices residing in the last two dimensions.- Returns
det – The determinant(s) of matrix. If
matrix.shape == (3, 3)
, the determinant will be returned as a scalar. Ifmatrix.shape == (..., 3, 3)
, the determinants will be returned as anumpy.ndarray
of shape(...,)
and dtypenumpy.float64
.- Return type
- Raises
ValueError: – If matrix has less than two dimensions or its last two dimensions are not of shape
(3, 3)
.
New in version 0.20.0.
- MDAnalysis.lib.mdamath.triclinic_box(x, y, z)[source]
Convert the three triclinic box vectors to
[lx, ly, lz, alpha, beta, gamma]
.If the resulting box is invalid, i.e., any box length is zero or negative, or any angle is outside the open interval
(0, 180)
, a zero vector will be returned.All angles are in degrees and defined as follows:
alpha = angle(y,z)
beta = angle(x,z)
gamma = angle(x,y)
- Parameters
x (array_like) – Array of shape
(3,)
representing the first box vectory (array_like) – Array of shape
(3,)
representing the second box vectorz (array_like) – Array of shape
(3,)
representing the third box vector
- Returns
A numpy array of shape
(6,)
and dtypenp.float32
providing the unitcell dimensions in the same format as returned byMDAnalysis.coordinates.base.Timestep.dimensions
:[lx, ly, lz, alpha, beta, gamma]
.Invalid boxes are returned as a zero vector.
- Return type
Note
Definition of angles: http://en.wikipedia.org/wiki/Lattice_constant
See also
Changed in version 0.20.0: Calculations are performed in double precision and invalid box vectors result in an all-zero box.
- MDAnalysis.lib.mdamath.triclinic_vectors(dimensions, dtype=<class 'numpy.float32'>)[source]
Convert
[lx, ly, lz, alpha, beta, gamma]
to a triclinic matrix representation.Original code by Tsjerk Wassenaar posted on the Gromacs mailinglist.
If dimensions indicates a non-periodic system (i.e., all lengths are zero), zero vectors are returned. The same applies for invalid dimensions, i.e., any box length is zero or negative, or any angle is outside the open interval
(0, 180)
.- Parameters
dimensions (array_like) –
Unitcell dimensions provided in the same format as returned by
MDAnalysis.coordinates.base.Timestep.dimensions
:[lx, ly, lz, alpha, beta, gamma]
.dtype (numpy.dtype) – The data type of the returned box matrix.
- Returns
box_matrix – A numpy array of shape
(3, 3)
and dtype dtype, withbox_matrix[0]
containing the first,box_matrix[1]
the second, andbox_matrix[2]
the third box vector.- Return type
Notes
The first vector is guaranteed to point along the x-axis, i.e., it has the form
(lx, 0, 0)
.The second vector is guaranteed to lie in the x/y-plane, i.e., its z-component is guaranteed to be zero.
If any box length is negative or zero, or if any box angle is zero, the box is treated as invalid and an all-zero-matrix is returned.
Changed in version 0.7.6: Null-vectors are returned for non-periodic (or missing) unit cell.
Changed in version 0.20.0: * Calculations are performed in double precision and zero vectors are also returned for invalid boxes. * Added optional output dtype parameter.
- MDAnalysis.lib.mdamath.box_volume(dimensions)[source]
Return the volume of the unitcell described by dimensions.
The volume is computed as the product of the box matrix trace, with the matrix obtained from
triclinic_vectors()
. If the box is invalid, i.e., any box length is zero or negative, or any angle is outside the open interval(0, 180)
, the resulting volume will be zero.- Parameters
dimensions (array_like) –
Unitcell dimensions provided in the same format as returned by
MDAnalysis.coordinates.base.Timestep.dimensions
:[lx, ly, lz, alpha, beta, gamma]
.- Returns
volume – The volume of the unitcell. Will be zero for invalid boxes.
- Return type
Changed in version 0.20.0: Calculations are performed in double precision and zero is returned for invalid dimensions.
13.2.7.2. Connectivity
- MDAnalysis.lib.mdamath.make_whole()
Move all atoms in a single molecule so that bonds don’t split over images.
This function is most useful when atoms have been packed into the primary unit cell, causing breaks mid molecule, with the molecule then appearing on either side of the unit cell. This is problematic for operations such as calculating the center of mass of the molecule.
+-----------+ +-----------+ | | | | | 6 3 | | 3 | 6 | ! ! | | ! | ! |-5-8 1-2-| -> | 1-2-|-5-8 | ! ! | | ! | ! | 7 4 | | 4 | 7 | | | | +-----------+ +-----------+
- Parameters
atomgroup (AtomGroup) – The
MDAnalysis.core.groups.AtomGroup
to work with. The positions of this are modified in place. All these atoms must belong to the same molecule or fragment.reference_atom (
Atom
) – The atom around which all other atoms will be moved. Defaults to atom 0 in the atomgroup.inplace (bool, optional) – If
True
, coordinates are modified in place.
- Returns
coords – The unwrapped atom coordinates.
- Return type
- Raises
NoDataError – There are no bonds present. (See
guess_bonds()
)ValueError – The algorithm fails to work. This is usually caused by the atomgroup not being a single fragment. (ie the molecule can’t be traversed by following bonds)
Example
Make fragments whole:
from MDAnalysis.lib.mdamath import make_whole # This algorithm requires bonds, these can be guessed! u = mda.Universe(......, guess_bonds=True) # MDAnalysis can split AtomGroups into their fragments # based on bonding information. # Note that this function will only handle a single fragment # at a time, necessitating a loop. for frag in u.atoms.fragments: make_whole(frag)
Alternatively, to keep a single atom in place as the anchor:
# This will mean that atomgroup[10] will NOT get moved, # and all other atoms will move (if necessary). make_whole(atomgroup, reference_atom=atomgroup[10])
New in version 0.11.0.
Changed in version 0.20.0: Inplace-modification of atom positions is now optional, and positions are returned as a numpy array.
- MDAnalysis.lib.mdamath.find_fragments()
Calculate distinct fragments from nodes (atom indices) and edges (pairs of atom indices).
- Parameters
atoms (array_like) – 1-D Array of atom indices (dtype will be converted to
numpy.int64
internally)bonds (array_like) – 2-D array of bonds (dtype will be converted to
numpy.int32
internally), wherebonds[i, 0]
andbonds[i, 1]
are the indices of atoms connected by thei
-th bond. Any bonds referring to atom indices not in atoms will be ignored.
- Returns
fragments (list) – List of arrays, each containing the atom indices of a fragment.
.. versionaddded:: 0.19.0
New in version 0.11.0.