xrt (XRayTracer)




GHver GHdate


PPver PPdate version history, plans




Konstantin Klementiev (MAX IV Laboratory), Roman Chernikov (Canadian Light Source)

Package xrt is a python software library for ray tracing and wave propagation in x-ray regime. It is primarily meant for modeling synchrotron sources, beamlines and beamline elements. Includes a GUI for creating a beamline and interactively viewing it in 3D.

 Intensity of undulator radiation on a transverse flat screen. The long axis is energy. One can observe two harmonics crossing the central line: the one at 2400 eV is odd and the one at 4800 eV is even. Intensity of undulator radiation on a transverse flat screen. The long axis is energy. One can observe two harmonics crossing the central line: the one at 2400 eV is odd and the one at 4800 eV is even.

 Absorbed power and power density on a mirror at varying pitch angle. Absorbed power and power density on a mirror at varying pitch angle.

 Vortex beam with l=1 and p=1. Colored by phase. Vortex beam with l=1 and p=1. Colored by phase.


xrtQook – a GUI for creating a beamline

The main interface to xrt is through a python script. Many examples of such scripts can be found in the supplied folders examples and tests. The script imports the modules of xrt, instantiates beamline parts, such as synchrotron or geometric sources, various optical elements, apertures and screens, specifies required materials for reflection, refraction or diffraction, defines plots and sets job parameters.

The Qt tool xrtQook takes these ingredients as GUI elements and prepares a ready to use script that can be run within the tool itself or in an external Python context. xrtQook has a parallelly updated help panel that provides a complete list of parameters for the used objects. xrtQook writes/reads the recipes of beamlines into/from xml files.

In the present version, xrtQook does not provide automated generation of scans and does not create wave propagation sequences. For these two tasks, the corresponding script parts have to be written manually based on the supplied examples and the present documentation.

See a brief tutorial for xrtQook.

 A view of xrtQook with an empty beamline tree on the left and a help panel on the right. A view of xrtQook with an empty beamline tree on the left and a help panel on the right.

Features of xrt

  • Rays and waves. Classical ray tracing and wave propagation via Kirchhoff integrals, also freely intermixed. No further approximations, such as thin lens or paraxial. The optical surfaces may have figure errors, analytical or measured. In wave propagation, partially coherent radiation is treated by incoherent addition of coherently diffracted fields generated per electron. Propagation of individual coherent source modes is possible as waves, hybrid waves (i.e. partially as rays and then as waves) and only rays.

  • Publication quality graphics. 1D and 2D position histograms are simultaneously coded by hue and brightness. Typically, colors represent energy and brightness represents beam intensity. The user may select other quantities to be encoded by colors: angular and positional distributions, various polarization properties, beam categories, number of reflections, incidence angle etc. Brightness can also encode partial flux for a selected polarization and incident or absorbed power. Publication quality plots are provided by matplotlib with image formats PNG, PostScript, PDF and SVG.

  • Unlimited number of rays. The colored histograms are cumulative. The accumulation can be stopped and resumed.

  • Parallel execution. xrt can be run in parallel in several threads or processes (can be opted), which accelerates the execution on multi-core computers. Alternatively, xrt can use the power of GPUs via OpenCL for running special tasks such as the calculation of an undulator source or performing wave propagation.

  • Scripting in Python. xrt can be run within Python scripts to generate a series of images under changing geometrical or physical parameters. The image brightness and 1D histograms can be normalized to the global maximum throughout the series.

  • Synchrotron sources. Bending magnet, wiggler, undulator and elliptic undulator are calculated internally within xrt. Please look the section Comparison of synchrotron source codes for the comparison with other popular codes. If the photon source is one of the synchrotron sources, the total flux in the beam is reported not just in number of rays but in physical units of ph/s. The total power or absorbed power can be opted instead of flux and is reported in W. The power density can be visualized by isolines. The magnetic gap of undulators can be tapered. Undulators can be calculated in near field. Custom magnetic field is also possible. Undulators can be calculated on GPU, with a high gain in computation speed, which is important for tapering and near field calculations.

  • Shapes. There are several predefined shapes of optical elements implemented as python classes. The python inheritance mechanism simplifies creation of other shapes: the user specifies methods for surface height and surface normal. The surface and the normal are defined either in local Cartesian coordinates or in user-defined parametric coordinates. Parametric representation enables closed shapes such as capillaries or wave guides. It also enables exact solutions for complex shapes (e.g. a logarithmic spiral or an ellipsoid) without any expansion. The methods of finding the intersections of rays with the surface are very robust and can cope with pathological cases such as sharp surface kinks. Notice that the search for intersection points does not involve any approximation and has only numerical inaccuracy which is set by default as 1 fm. Any surface can be combined with a (differently and variably oriented) crystal structure and/or (variable) grating vector. Surfaces can be faceted.

  • Energy dispersive elements. Implemented are crystals in dynamical diffraction, gratings (also with efficiency calculations), Fresnel zone plates, Bragg-Fresnel optics and multilayers in dynamical diffraction. Crystals can work in Bragg or Laue cases, in reflection or in transmission. The two-field polarization phenomena are fully preserved, also within the Darwin diffraction plateau, thus enabling the ray tracing of crystal-based phase retarders.

  • Materials. The material properties are incorporated using three different tabulations of the scattering factors, with differently wide and differently dense energy meshes. Refractive index and absorption coefficient are calculated from the scattering factors. Two-surface bodies, such as plates or refractive lenses, are treated with both refraction and absorption.

  • Multiple reflections. xrt can trace multiple reflections in a single optical element. This is useful, for example in ‘whispering gallery’ optics or in Montel or Wolter mirrors.

  • Non-sequential optics. xrt can trace non-sequential optics where different parts of the incoming beam meet different surfaces. Examples of such optics are poly-capillaries and Wolter mirrors.

  • Singular optics. xrt correctly propagates vortex beams, which can be used for studying the creation of vortex beams by transmissive or reflective optics.

  • Global coordinate system. The optical elements are positioned in a global coordinate system. This is convenient for modeling a real synchrotron beamline. The coordinates in this system can be directly taken from a CAD library. The optical surfaces are defined in their local systems for the user’s convenience.

  • Beam categories. xrt discriminates rays by several categories: good, out, over and dead. This distinction simplifies the adjustment of entrance and exit slits. An alarm is triggered if the fraction of dead rays exceeds a specified level.

  • Portability. xrt runs on Windows and Unix-like platforms, wherever you can run python.

  • Examples. xrt comes with many examples. See the galleries, the links are at the top bar.

  • Tests. xrt comes with the tests of its several main aspects. See the folder tests and also this documentation for the tests of undulators, optical elements, materials and wave propagation.


numpy, scipy and matplotlib are required. If you use OpenCL for calculations on GPU or CPU, you need AMD/NVIDIA drivers, Intel CPU only OpenCL runtime (these are search key words), pytools and pyopencl. PyQt4 or PyQt5 are needed for xrtQook. Spyder (as library of Spyder IDE) is highly recommended for nicer view of xrtQook. OpenGL is required for xrtGlow.

See detailed instructions for installing dependencies.

Get xrt

xrt is available as source distribution from pypi.python.org and from GitHub. The distribution archive also includes tests and examples. The complete documentation is available online on Read the Docs and offline as zip file on GitHub.


Unzip the .zip file into a suitable directory and use sys.path.append(path-to-xrt) in your script. You can also install xrt to the standard location by running python setup.py install from the directory where you have unzipped the archive, which is less convenient if you try different versions of xrt and/or different versions of python. Note that python-64-bit is by ~20% faster than the 32-bit version (tested with WinPython). Also consider Speed tests for making choice between Windows and Linux and between Python 2 and Python 3.

Get help

For getting help and/or reporting a bug please use GitHub xrt Issues.

Citing xrt

Please cite xrt as: K. Klementiev and R. Chernikov, “Powerful scriptable ray tracing package xrt”, Proc. SPIE 9209, Advances in Computational Methods for X-Ray Optics III, 92090A (2014); doi:10.1117/12.2061400. Online documentation on xrt.readthedocs.io; doi:10.5281/zenodo.1252468.

For the new capabilities: 1) Calculations of elastically deformed crystals on GPUs, 2) Mode analysis and mode propagation, 3) Modelling OAM (Orbital Angular Momentum) beams, please cite: K. Klementiev & R. Chernikov, “New Features of xrt: Bent Crystals, Coherent Modes, Waves with OAM” Synchrotron Radiation News (2023) 36:5, 23-27; doi:10.1080/08940886.2023.2274735.


  • Josep Nicolás and Jordi Juanhuix (synchrotron Alba) are acknowledged for discussion and for their Matlab codes used as examples at early stages of the project.

  • Andrew Geondzhian and Victoria Kabanova (summer students at DESY) are acknowledged for their help in coding the classes of synchrotron sources.

  • Rami Sankari and Alexei Preobrajenski (MAX IV Laboratory) are thanked for discussion, testing and comparing with external codes.

  • Hasan Yavaş, Jozef Bednarčik, Dmitri Novikov and Alexey Zozulya (DESY Photon Science) are acknowledged for supplied cases.

  • Hamed Tarawneh (MAX IV Laboratory) has initiated the custom field undulator project and supplied tables of magnetic field.

  • Bernard Kozioziemski (LLNL) has pointed to the importance of considering the total absorption cross-section (not just photoelectric absorption) in the cases of light materials at high energy. He also did deep tests of the mosaic crystal model.

  • Emilio Heredia (CLS) is thanked for valuable suggestions on GUI.

  • Tim May (CLS) is acknowledged for inspiring the custom magnetic field calculation development, providing field tables and comparing the results with other codes.

  • Louisa Pickworth (MAX IV Laboratory) has provided curves from other codes for testing reflectivity/transmittivity of multilayers.

User’s Guide