Introduction to WebbPSF for Roman

WebbPSF provides a customizable interface to perform point-spread function (PSF) simulations and calculations for the Wide-Field Instrument (WFI) planned for the Nancy Grace Roman Space Telescope. A simulated PSF provides a useful tool for predicting the observatory's performance at a particular combination of wavelength (or bandpass), field position, and pixel scale. PSFs are also an important input to simulate astronomical scenes. For example, WebbPSF-simulated PSFs are used by the Pandeia and STIPS toolkits developed at the Space Telescope Science Institute (STScI).

Roman Simulation Tools

Point-spread function for a G0V star
Simulated point-spread functions of a GOV star observed in eight filters with the Wide Field Instrument of the Roman Space Telescope. Intensity is shown in log scale. Axes are labeled in arcseconds.


PSFs are simulated in the far-field limit (Fraunhofer domain) using the same optical simulation library as the existing WebbPSF software. WebbPSF was developed to simulate the James Webb Space Telescope (JWST) instrument PSFs, and its accuracy has been checked against ground test data for JWST (see the references listed below). The WebbPSF WFI model is based on the Phase C payload design, and incorporates information from GSFC Roman Space Telescope Instrument Reference Information files.

WebbPSF allows users to calculate PSFs both in the bandpasses defined in the GSFC Roman instrument reference data and for the monochromatic case (which can be useful as an input to other calculations). For broadband PSFs, WebbPSF allows the user to select an input spectrum (e.g., a stellar spectral type or a galaxy spectral energy distribution), which then weights the individual monochromatic components of the PSF appropriately.

The software provides several built-in analysis tools to compute a radial profile, an encircled energy curve, or the difference image of two PSFs. All of these analysis tools work with standard FITS files with appropriate header keywords, and the PSF calculation results from WebbPSF can be written directly to FITS files for export to other tools.

Compared to the current WebbPSF instrument models for JWST, WebbPSF adds support for field-dependent PSF aberrations both within a single detector and among the 18 detectors in the WFI focal plane. These are modeled as Zernike coefficients, using the instrument reference data from the Phase C payload design.


Getting Started with WebbPSF

  • The WebbPSF package includes both JWST and Roman instrument models and is available for download and installation. The WebbPSF installation documentation contains instructions and options for installation, that do not depend on the specific functionality (JWST vs Roman).
  • Tutorial Jupyter notebooks are available on GitHub to aid users interested in getting started with STScI’s simulation tools for Roman, including WebbPSF.
Example calculation showing the assumed high-spatial-frequency wavefront error map
Example calculation showing the assumed high-spatial-frequency wavefront error map due to the telescope assembly, the addition of field dependent aberrations in the instrument focal plane, and the calculated PSF (before sampling on the detector pixels).

Citing WebbPSF

A technical report (WFIRST-STScI-TR1703: WebbPSF for WFIRST) by Long and Perrin discusses the assumptions underlying WebbPSF, and compares its model PSFs to those calculated with alternative tools. Users are also encouraged to cite one of the following publications covering WebbPSF's JWST functionality:

"Updated point spread function simulations for JWST with WebbPSF"
Perrin et al., 2014. Proc. SPIE. Vol. 9143.

"Simulating point spread functions for the James Webb Space Telescope with WebbPSF"
Perrin et al., 2012. Proc. SPIE. Vol. 8442.


Assumptions and Caveats

The Roman functionality of WebbPSF is still in continued development. The present implementation is available to the community (1.0.0).

The Roman mission is currently in development, and observatory designs continue to evolve. Hence, WebbPSF simulations are the best effort to provide a reasonable description of the actual observatory.

At this stage, the model for the WFI makes several approximations beyond the basic assumption of far-field diffraction-based PSF calculation:

  • The filter transmission for broadband PSF calculations is assumed to be flat across the filter band pass.
  • No attempt is made to model detector effects such as inter-pixel capacitance or noise characteristics. These issues can be assessed instead with the Pandeia software.
  • Aberrations are represented as Zernike polynomials from Z4 to Z22. Since measurements of the wavefront error for the Roman primary mirror are not yet available, higher-order aberrations beyond Z22 are approximated by using higher spatial frequency wavefront error content based on the Optical Path Difference (OPD) map across the Hubble Space Telescope primary mirror.
  • The WFI model selects the pupil shape best represents the F158, F184 and F146 filters and the grism which are mounted with proximate cold pupil masks. For custom bandpasses or monochromatic calculations, the unmasked pupil is used unless the maximum wavelength requested for the calculation is greater than 1.454 µm.
  • WebbPSF does not yet model distortions of the apparent pupil shape as seen from the detector.
  • The astrometry of point sources is only accurate to the +/- 0.5 pix.
  • Images are generated in units of electrons.
  • The photometric accuracy is undergoing further validation.
  • Pixel effects (such as intrapixel sensitivity, interpixel capacitance; pixel response functions, and/or non-linearities) are not addressed.

The following caveats should be heeded:

  • If the oversample parameter is set to an even number, then the size of final image will in turn be an even number regardless of the number of pixels requested in the output model PSF. Consequently, the PSF will be centered on the intersection of a 2x2 grid of pixels at the image's center. If users require a PSF that is centered on a pixel (which would imply an image that has an odd size) and is oversampled, then they must employ the following work-around. Once the WFI object is instantiated, users must manually set the detector pixelscale as an integer fraction of the native pixel scale, correspondingly increase the number of pixels in the fov, and keep the oversample parameter as 1. For example, if one requires a PSF that is 91 (native) pixels in size with an oversampling of 4, then the pixel scale, fov, and oversample should be specified as 0.11/4, 361, and 1, respectively.
  • Very large PSFs (>~500 pixels in radius) will suffer from numerical artifacts that manifest as 'inward facing spikes.'
  • Note that WebbPSF pixel positions are zero-indexed.


Future Improvements

We are planning many additions and improvements for future WebbPSF versions, including the following:

  • Improvements to the graphical notebook interface and additional documentation.
  • Modeling of the distortion of the apparent pupil shape as seen from different points in the field of view.
  • Better filter transmission curves and high spatial frequency wavefront error maps, once these have been measured and released by the project.
  • Other fixes and features based on community feedback.


Credits, Acknowledgements, and Feedback

WebbPSF is developed by Marshall Perrin, Robel Geda, and collaborators in support of STScI's involvement in both the JWST and Roman missions. Software development takes place on GitHub. Contributions of bug reports or code are welcome!

Users are encouraged to address questions, suggestions, and bug reports to with "WebbPSF" in the subject line, where they will be directed to the appropriate members of the WebbPSF team at STScI.


Nancy Grace Roman Space Telescope insignia

The NASA Nancy Grace Roman Space Telescope is managed by NASA/GSFC with participation of STScI, Caltech/IPAC, and NASA/JPL.

Contact the Roman Team