The XSPEC variance models calculated so far can be found in a tarball below. Models are calculated with the Monte-Carlo method described in Parker et al 2019. These models are simple, cover a limited parameter space, and rely on basic assumptions of variability. Use at your own risk! We strongly suggest reading the guidelines below, in full, before use.

Guidelines for use:

  • For models with a normalisation, fix this at 1, otherwise the results are not valid

  • No Galactic absorption is required, or any other constant multiplicative component.

  • Do not use these models with count spectra. That’s not what they’re for.

  • Because the models are relatively unsophisticated compared to conventional X-ray models, we recommend adding a small 1-2% systematic error. This is probably not necessary for low quality RMS spectra.

  • Note that some versions of xspec have a bug when adding systematic error that may cause them to not fit. This can be avoided by running an initial fit without systematic error, before adding it in and re-fitting.

  • For models derived from Spex models (for example, fvar_ufo), the unit convention follows that of Spex (SI units) rather than xspec (cgs).

  • When you use these models, please cite Parker et al. 2020.

New models:
New models will be added as we generate them. If you need a specific model that we do not provide, please get in touch.



The SPEX spectral fitting package (Kaastra et al., 1996) has very precise models for photoionised emission and absorption that are not available natively for XSPEC. I have generated XSPEC tables for some of the XABS photoionised absorption model, and the emitted spectrum from the PION photoionised emission/absorption model.

These models are evaluated at grid points and interpolated, so they are by definition less accurate than the native SPEX  versions. Use at your own risk!

If you do use these models, please cite:

Kaastra, Mewe and Nieuwenhuijzen, 1996 (SPEX)
Miller et al., 2015 and Mehdipour et al., 2016 (PION)
Steenbrugge et al., 2003 (XABS)
Parker et al., 2019 (XSPEC table implementation)


Note that, depending on your needs, Jeremy Sanders' more sophisticated implementation may be a better tool: github link


I have generated some custom reflection models using the Reflionx code (Ross & Fabian, 2005). This code is relatively fast, so custom models can be developed within a week or so. If you would like such a model, please contact me


reflionx_hc: reflionx with a high energy cutoff, ideal for use with NuSTAR data (Tomsick et al., 2014).

reflionx_bb: uses an input black body spectrum, rather than the standard powerlaw (King et al. 2016).

reflionx_nthcomp: uses a Comptonised input spectrum, using the NTHComp model (Zdziarski, Johnson & Magdziarz 1996, Zycki, Done & Smith 1999).