c2x - converts various crystal formats including density grid data


c2x [-OPTIONS] [--FORMAT] [--OPERATION] infile [outfile]


c2x converts primarily a CASTEP .check file to various output formats, additionally extracting densities (charge, spin, band or psi) and forces. It can also read CASTEP .cell files and PDB files, Onetep .dat files, and several input and output files from Abinit, Quantum Espresso and VASP, with some support for Siesta too. It is a sort of Babel with support for gridded data and .check files, and the ability to transform cells and perform simple analysis (integration, interpolation, dipole moment calculation, band parity identification).

It may have been compiled to give access to symmetry functions from spglib too.

An input file whose name ends ’.pdb’ is assumed to be in pdb format, ending in ’.cif’ is assumed to be in cif format, ending in ’.res’ is assumed to be in shelx97, ending ’.cub’ or ’.cube’ is assumed to be in cube format, ending ’.in’ is assumed to be an Abinit or Quantum Espresso input file, ending ’.xml’ is assumed to be a Quantum Espresso output file, ending in ’.xsf’ is assumed to be in xsf format, and ending in ’.xv’ or ’XV’ is assumed to be in Siesta’s XV format. Input files ending CHG, CHGCAR, POSCAR, CONTCAR, LOCPOT or WAVECAR are assumed to be in VASP 5.x format. Input files ending in DEN, POT or WFK are assumed to be in Abinit format. Otherwise, if the first byte is either zero, 10 or 30 it is assumed to be a .check file, else it is assumed to be a .cell file. It can also read .orbitals files (which are identical to .check files in format), and .castep_bin files (which lack wavefunctions). Furthermore, it can read .chdiff files and .cst_esp files. In these cases it needs a .cell or a .check file as well in order to obtain unit cell information.



rotate as though outputing in abc format, i.e. place a along x axis, b in xy plane, and abc form a right hand set. Useful if one wants a dx file consistent with a pdb file. See also -3.


accumulate (sum) bands requested by -b= or -B=.


include specified bands as psi (real).


include specified bands as densities (psi*conjg(psi)).


include charge density (units electrons per cubic Angstrom).


evaluate next arg with arithmetic parser and exit.


report internal conversion constants and exit.


equivalent to --simpson --tricubic


find "compact" (near-cubic) set of cell vectors.


read also a corresponding .chdiff file, and output its contents. The filename given must still be that of a .cell or .check file, as a .chdiff file contains no axes.


replace a vector dataset with its (scalar) divergence.


if charge density read (-c), calculate dipole moment about fractional co-ordinates x,y,z, or 0.5,0.5,0.5 if co-ordinates not given. Assumes density has been read as eA^-3.


as above, but also report post-hoc energy correction for slab geometry for the a axis being the non-periodic axis. Valid values of a: a, b and c.


as above, but also report post-hoc energy correction for a molecule in a cubic box, or for a molecule in a tetragonal box if dipole moment is parallel to c.


read also a corresponding .cst_esp file, and output its contents. The filename given must still be that of a .cell or .check file, as a .cst_esp file contains no axes.


set symmetry tolerance to given number of Angstroms.


calculate electrostatic potential, assuming that an electron density has been read (implies -c). Ions are treated as Gaussian blobs of charge of extent exp(-mu^2r^2). If the ionic charge differs from the atomic number, a further localised smoothing of the atomic potential occurs, unless the first character after the = is -. With -E=0, a bare Coulomb potential is used.


calculate first failure start of k-point set.


calculate first Brillouin zone


use FFT interpolation to move gridded data between different unit cells, and with -z.


list supported formats.


extract single frame from series of time steps. Frames are numbered from zero, and negative numbers represent offsets from the end of the sequence, so --frame=-1 will extract the final frame.


calculate the potential, then its gradient, to give an electric field.


print band gap in eV.


replace a scalar dataset with its (vector) gradient.


replace a scalar dataset with the result of the Laplacian operator.


shift atoms by half a grid cell. For use with xplor data format, see below.


Fourier interpolate onto specified grid size. New grid may be coarser or finer than original. Any dimension given as zero is replaced by old grid size. If reading wavefunction, any grid truncation is done after transforming back to real space and converting to density etc.


calculate irreducible Brillouin zone.


report whether bands have inversion, and parity under inversion. If combined with -b or -B, the last range given is used.


include given kpoints for bands (default range is 1).


if k-points are to be included in a .cell file, explicitly list them rather than using the MP generation parameters.


output in abc format assuming that abc describes a left-handed set of axes. Do not use this unless you understand why you should not!


assume input is molecule, not crystal. Try to avoid outputing a cell, shift if some co-ordinates are negative, or if a,b,c given shift by those numbers of FFT grid cells.


assume input is molecule, not crystal. Try to avoid outputing a cell, shift atoms by fractional co-ordinates given.


discard symmetry information, and, if output is XSF, discard forces. Give twice to discard k-points too.


normalise by reducing fractional coords to 0<=x<1. In conjunction with -m, do output a cell.


print band occupancies and evalues to stderr.


find primitive cell with own internal algorithm, not spglib.


output data as line of nn points from p1 to p2. Express p’s as either fractional co-ordinates in the form (x,y,z), or an atom position as, e.g., Si3 for silicon atom number 3, or simply Si for the first Si atom. Using a 0 (zero) for a p is equivalent to (0,0,0), and the three cell axes "(0,0,0):(1,0,0):ngx+1" (etc.) can be specified as a, b and c. Each co-ordinate is passed to the arithmetic parser, so 1/3 is acceptable for a third, etc.


output data with cylindrical averaging. p1 is the centre of the cylinder, rl (literal "r", followed by a length, suffixed by "B" if Bohr) is the radius, and nn the number of points. The axis of the cylinder must be the c axis, and alpha and beta must be 90 degrees. Data are averaged over c and theta. Append "w" to number of points to weight samples by two pi times radius, or "a" to weight and accumulate.


output data with spherical averaging. p1 is the centre of the sphere, Rl (literal "R", followed by a length, suffixed by "B" if Bohr) is the radius, and nn the number of points output. The number of points for sampling around the spherical surface is chosen to give a similar point separation to that along the line. A length of zero will set the length to the maximum possible given the periodicity. Data are averaged over theta and phi. Append "w" to number of points to weight samples by four pi times radius squared, or "a" to weight and accumulate.


calculate post hoc energy correction for charged isolated system. Implies -c.


calculate post hoc energy correction for charged 2D system. The axis given (a, b or c) is the aperiodic one. Implies -c.


sort the atoms on output in descending atomic order


sort the atoms on output in ascending atomic order


only reduced (symmetry inequivalent) atoms in cif output, or with the --fbz or --ibz options, reduce k-points to the relevant zone, adjusting weights appropriately.


don’t attempt to rescale densities, but output them raw. Charge density becomes electrons per unit cell if reading from Castep, for instance. Also do not attempt to adjust radius to maintain bond length on nanotube creation.


rescale grid data by factor x, not whatever factor (if any) would normally be used. If the factor is suffixed by an "x", do include c2x’s usual conversion factor too.


include spin density (scalar or vector).


use Simpson’s rule for accumulating/integrating, not trapezium rule.


list symmetry elements found in input, without calling SPGlib.


include specified spins or spinors for bands (default range is -, and the spins are numbered 0 and 1).


rotate co-ordinate system so that the first vector becomes the second. First vector given in relative co-ordinates. If third axis given, it is used as the rotation axis. Else the rotation axis will be perpendicular to the two axes given.


when interpolating, using tricubic (spline) interpolation in place of trilinear interpolation. Slower, might not preserve monotonicity, but does keep first derivative continuous.


rotate co-ordinate system so that the first vector becomes the second. All vectors given in absolute co-ordinates. If third axis given, it is used as the rotation axis. Else the rotation axis will be perpendicular to the two axes given.


use atomic units (Bohrs) when writing .cell files and 1D data. Scale densities from A^-3 to Bohr^-3 when writing .cube files.


scale densities from Bohr^-3 to A^-3 when reading .cube files.


be verbose. Far too much output can be generated if specified more than twice.


convert vector field to "forces" on each atom. Use value of vector at atom, unless =r given, when average over a sphere of radius r (Angstroms, unless suffixed with B for Bohr). If i given, integrate rather than average. If f given, use finer grid for averages/integrals. The f may be repeated. If species given, limit "forces" to given atomic species. See also --cubic.


print version information. If preceeded by -v, also print internal conversion factors.


replace vector grid data by its scalar modulus.


weight bands by occupancies, or sqrt(occ) if not calculating density.


weight bands by k-point weight, but not occupancy.


calculate band widths naively. Makes sense only if bands do not cross.


weight bands by occupancies and k-point weight, or sqrt thereof if not calculating density.


expand unit cell to new cell specified in terms of the old cell axes. Each co-ordinate is passed to the arithmetic parser, so ’sqrt(3)/2’ etc is acceptable. Optionally also specify new grid size for gridded data, else this is calculated automatically.


expand cell with a trivial tiling.


expand unit cell to new cell specified in absolute co-ordinates. Optionally also specify new grid size for gridded data, else this is calculated automatically.


change given axis/axes to new length by inserting / removing vacuum around the origin. Removing non-existent vacuum will produce nonsense. Length may be suffixed with B (for Bohr) or nm.


make nanotube. The input cell must have c perpendicular to the ab plane, and c as the nonperiodic direction of the sheet to be rolled. The circumference is then defined by the vector i*a+j*b. The vector along the tube’s length found automatically, and the size of the cell perpendicular to the tube’s length is given by the optional parameter x, which may be suffixed with B (for Bohr) or nm.


print to stdout data at given point, and set output type to null. For specification of p1, see -P= option.


ditto, but assume that data represents an electron density in A^-3, and also output Perdew Zunger 81 XC energy.


when moving from a left hand set of axes to a right hand set, rather than exchanging the 2nd and 3rd axes, preserve the 3rd and exchange the 1st and 2nd. This transformation is required if the input is cartesian and left handed, and an abc output is requested. Specifying this flag twice will cause the 1st and 3rd axes to be exchanged.


The following options expect multiple input files to be given, and perform the specified operation.


Add datasets element-wise.


Subtract datasets element-wise.


Multiply datasets element-wise. Although it is assumed that one dataset will be a mask of ones and zeros, it need not be so.


Merge datasets. The expected use is merging an atoms-only format with a density-only format to create an output containing both atoms and density.


Alternative for --mask.


Alternative for --diff.


Alternative for --add.

When merging, if files contain conflicting data, the one on the right usually has precidence.

When performing operations on grids, the grids must be the same size. The use of -i may assist. In all cases, the cells must be the same.


The following output formats are recognised. (See the output of the --formats option for the complete list. Most are also recognised for input provided that their filenames have the expected suffix.)

Abinit .abi file (for Abinit version 9 and beyond).


Abinit .in file. The output is insufficient to be a valid input file to Abinit, but can easily be made so.


CASTEP .bands file, no sorting of bands.


XCrysDen / FermiSurfer file for plotting Fermi surfaces. A symmetry-reduced kpoint set will be expanded. (Not accepted as input.)


CCP4 density map format. Note no atomic positions can be recorded in this format, and c2x will always produce a right-hand set of axes unless the input is a lhs and the option -L is given.


CASTEP .cell, cartesian cell, fractional co-ordinates.


CASTEP .cell, abc cell, fractional co-ordinates.


CASTEP .cell, cartesian cell, absolute co-ordinates.


CASTEP .cell, abc cell, absolute co-ordinates.


VASP 5.x chgcar output.


a very basic and rigid format which may be compatible with some CIF-reading software.


Chemical Markup Language.


Gaussian cube. Atoms and at most one data set.


Data Explorer. Data set only.


CASTEP formatted density


Elk elk.in format.


Siesta. If a density has been read, a corresponding .RHO file will be written.


An ASCII wavefunction coefficient representation


The same, sorted by |g|


Gnuplot command file for 1D data.


Jmol output for Brillouin zones (with --fbz or --ibz only).


Numpy array, single dataset, as doubles with -15, else single precision.


Null output. Throw away all output, but still write some useful information the input to stderr.


Onetep .dat, very similar to .cell. Also one_abc, one_abs, and one_abc_abs.




PDB, but label the atoms with element symbol and number within that species, e.g. C8, H24, Ca2, rather than just with element symbol. The whole string can contain no more than four characters, so * is used for the numeric part if it would not otherwise fit.


a python dictionary, compatible with the Atoms data structure from ASE.


a python ASE Atoms data structure.


Quantum Espresso. Non colinear spins not supported.


Ditto, atoms in fractional co-ordinates.


a subset of the SHELX97 format.


VASP 5.x output (poscar or chg).


raw list of vertices for Brillouin zones (with --fbz or --ibz only). Also --verts_frac for vertices in fractional coordinates.


VMD output for Brillouin zones (with --fbz or --ibz only).


Xplor format. Data set only. The grid used in this format is offset by half a grid cell compared to Castep, and as interpolating is inexact, this program does not in this case. Also the grid axes are described in terms of a, b, c, alpha, beta, gamma, so information about orientation in space is lost. To produce a compatible pdb file of atomic co-ordinates, specify -Hc when creating the pdb file.


XCrysDen format. Default. The only format in which multiple data sets are supported.


Siesta’s .XV format (positions only, velocities written as zero). Will also write a .RHO file if grid data have been read. On reading, will also read a FORCE_STRESS if present. Note that FORCE_STRESS has no prefix.


XYZ format. Atoms only, no unit cell.

Where a range is required, it can be specified as a single integer, two integers separated by a hyphen (all integers in the given range), or a comma-separated list of any of these elements. Only for the xsf output format is a range including more than a single integer meaningful.


If c2x has been compiled with spglib, the following spglib operations are available.


call spg_get_dataset() and report international symbol


call spg_get_dataset() and list symmetry ops


call spg_get_dataset() followed by spg_get_pointgroup()


call spg_find_primitive(), equivalent to spg_standardize_cell(to_primitive=1, no_idealize=0). This may rotate the cell to a standardised orientation.


call spg_standardize_cell(to_primitive=1, no_idealize=1), so primitive no rotation


call spg_refine_cell()


call spg_get_schoenflies()


call spg_standardize_cell() then expand back to a snapped version of the original cell


ditto, but include any translation introduced by spglib


call spg_standardize_cell(no_idealize=1)


call spg_standardize_cell(no_idealize=0)

--symmetry or --symm

call spg_get_dataset() and keep symmetry ops


For the pdb formats, just the unit cell and atomic positions are read or written. For the dx and xplor formats, just a single data set is written. For the Gaussian cube format atomic positions and at most one data set are recorded, and for the XCrysDen format the unit cell, atomic positions, forces, and any number of data sets are recorded.

When reading a .geom file and writing a format containing a single frame, the last frame is written, unless the --frame option is given.

When reporting symmetry operations, all co-ordinates are fractional.

Note that the pdb format offers a very low precision for storing co-ordinates, and, because it stores the unit cell in abc format, and the atoms in absolute coordinates, a rotation is likely to be required to place a on the cartesian x axis, etc. If so, it will be done automatically. The same is true for the abc varients of the cell format and for the Xplor and cif file formats. This rotation can be specified explicitly for other formats. Additionally the axes must form a right-handed set. If this is not the case, two axes will be interchanged. By default, b and c are exchanged, but the flag -3 will cause a and b to be exchanged instead.

The cif reader reads little more than c2x’s cif output. It is very basic, and will fail to read correctly a large number of valid cif files. There is currently no intention to produce a proper cif reader.

When outputting psi it is assumed that it is possible to make psi real by unwinding any phase produced by the k-point, and then multiplying all points by the same arbitrary complex constant. If this is not so, the band was probably nonsense anyway. The final choice of a factor of -1 is arbitrary. This scheme produces nonsense if one attempts to plot a degenerate band.

When doing the conversions resulting from -x, a new grid will be chosen of similar density to the old, and the data interpolated onto the new using trilinear interpolation. Extrapolating psis (rather than densities) is meaningless except at gamma, for the phase due to the k point is not considered.

Densities by default are in Angstroms**-3, and psis in Angstroms**-1.5, save that .RHO files are written in Bohr**-3 as expected.


To extract the charge density in xsf format

c2x -c input.check output.xsf

To extract the first four bands as psi at the second k-point in xsf format

c2x -b=1-4 -k=2 input.check output.xsf

To convert a check file to a pdb file

c2x --pdb input.check output.pdb

To convert a cell to something containing two repeat units in the a and b directions, and one in the c direction

c2x --cell -x=’(2,0,0)(0,2,0)(0,0,1)’ in.cell out.cell

or, from c2x version 2.30,

c2x --cell -x=2x2x1 in.cell out.cell

Assuming the above cell was a 3.5A cube, the same in absolute co-ordinates

c2x --cell -X=’(7,0,0)(0,7,0)(0,0,3.5)’ in.cell out.cell

To change a cell containing one layer of bulk in the c direction to one containing four layers, and sufficient vacuum to make a total length of 30A

c2x --cell -x=1x1x4 -Xc=30 in.cell out.cell


The following viewers have been used during the development of c2x: Avogadro, FermiSurfer, gabedit, Jmol, pymol, VESTA, VMD and XCrysDen.


None known.

Please report others to MJR.


If you wish to cite, please do so as "C2x: a tool for visualisation and input preparation for Castep and other electronic structure codes", MJ Rutter, Computer Physics Communications, vol 225 pages 174-179 (2018). http://dx.doi.org/10.1016/j.cpc.2017.12.008

Details of spglib can be found at https://atztogo.github.io/spglib/