A python library for performing reciprocal space reconstruction from single crystal x-ray measurements.
The package can be installed using pip:
pip install meerkat
On Windows we recommend to use virtual environments like anaconda which simplify python installation.
Some anaconda distributions fail on pip install meerkat while trying to compile h5py. In this case the following commands work:
pip install meerkat --no-deps
pip install fabio
Since anaconda comes with preinstalled h5py it is not required to reinstall it with pip, and the following command resolves all dependencies.
The reciprocal space reconstruction is based on the orientation matrix determined by XDS. Thus, in order to run the reconstruction, in addition to the diffraction frames, meerkat requires either XPARM.XDS or GXPARM.XDS.
The reconstruction can be run using the following python script:
from meerkat import reconstruct_data
#reconstruct dataset
reconstruct_data(filename_template='../frames/PdCPTN01002_%05i.cbf',
first_image=1,
last_image=3600,
reconstruct_in_orthonormal_basis=False,
maxind=[4,5,16], #the reconstruction will be made for h=-4...4, k=-5...5, l=-16...16
number_of_pixels=[801, 801, 801], #The resulting size of the array. Controls the step size
polarization_factor=0.5,
path_to_XPARM='/home/arkadiy/work/data/PdCPTN01002/xds',
output_filename='reconstruction.h5',
all_in_memory=False,
size_of_cache=100,
override=True)
The result is saved as an hdf5 file. The reconstruction is held in two datasets: rebinned_data and number_of_pixels_rebinned, the former is a corrected sum of intensities of reconstructed pixels, while the latter counts how many pixels were reconstructed. The scattering intensity can be obtained by dividing the two: rebinned_data[i,j,k]/number_of_pixels_rebinned[i,j,k].
In addition to the two datasets, the reconstruction file contains parameters of the reconstruction maxind, number_of_pixels, calculated step_size, and information from XDS files: unit_cell, space_group_nr and metric_tensor.
By default the reconstruction is performed in crystallographic coordinates. Such reconstructions can be easily symmetry-averaged. Also the numerical analysis of diffuse scattering is more straightforwardly performed in crystallographic coordinates (for example the program Yell uses such coordinates).
The downside of the crystallographic coordinates is that they are in general not orthorombic, which makes the reconstructions in such coordinates slightly more complicated to plot. If the reconstructions are required in orthonormal coordinates, this can be achieved by setting:
reconstruct_in_orthonormal_basis=True
If the reconstruction is performed in orthonormal basis, the new basis a*',b*',c*' is calculated from the crystal a*,b*,c* vectors. In the new basis a'* is parallel to a*, the b'* is in the plane spawned by a* and b*, and c'* is orthogonal to a'* and b'*.
The three dimensional arrays containing the reconstructed reciprocal space are typically large (~10Gb). We appreciate that not all computers might have enough operating memory to hold this datasets. Thanks to the hdf5, it is possible to use large arrays hosted on hard drive. In such case, only a small portion of the array will be cached in the operating memory. In order to turn on caching set parameter all_in_memory to False and define the size of the memory for cache.
all_in_memory = False
Such scheme is approximately three times slower, than holding all datasets in memory.