HCI PSF Reference Stars

JWST high-contrast imaging (HCI) often requires the observation of a nearby, unresolved reference star with similar spectro-photometric properties to the target of interest, to ensure effective PSF subtraction.

On this page

See also: NIRCam Coronagraphic PSF Estimation, NIRCam Coronagraphic Imaging Recommended Strategies, NIRISS AMI Recommended Strategies, NIRCam and MIRI Coronagraphy of the Beta Pictoris Debris Disk, NIRISS AMI Observations of Extrasolar Planets Around a Host StarNIRCam and MIRI Coronagraphy of HR8799 b

The baseline strategy for high-contrast imaging (HCI) with JWST includes the observation of a nearby star to generate an unresolved, high signal-to-noise (SNR) point spread function (PSF) to subtract from the science target, thereby reaching the highest possible contrast, with the goal of revealing faint astronomical signals surrounding the science target. This method is known as reference differential imaging (RDI).

Several factors can affect the quality of the RDI technique:

  • the PSF reference star is not single or is resolved (e.g., binary or disk)

  • the target acquisition (centering, in the case of a coronagraph) of the science target and PSF reference star differ (the PSF reference star is acquired at a different position and time)

  • position or thermally-induced wavefront drifts of the observatory resulting in the PSF reference star being no longer exactly the same as when the science target was acquired

  • the science target and PSF reference star differ in color or spectral energy distribution

The first factor can be addressed by selecting known good PSF reference stars, but this is not always trivial. The second factor can be addressed by using the small grid dither technique. The observer can minimize the impact of the latter two factors by (1) choosing a reference star in relative proximity to the science target (to mitigate thermal changes) and (2) by selecting a reference star that is spectro-photometrically similar to the science target. Choosing a nearby reference star also minimizes the telescope overheads (by reducing slew time). By including the science observation(s) and the PSF reference observation in a non-interruptible sequence, the visibility windows of the science and reference star must necessarily overlap at the time of the desired observation.

So how close in the spatial dimension must the science PSF star be to minimize thermal effects, and how close in spectral properties must they be for an acceptable match? There is no simple answer, but some guidelines may help:

Effect of spectral "mismatch"

See also: HCI Inner Working Angle

The spectral mismatch between a science target (hereafter “SCIENCE”) and its designated PSF reference star (hereafter “REFERENCE") has a stronger impact at shorter wavelengths and with wider filters. For a simple monochromatic case (narrowband filter in the continuum), when performing the PSF subtraction (SCIENCE − REFERENCE), one needs to account for the flux difference and photometrically rescale the REFERENCE. If the REFERENCE is fainter, the process of flux rescaling also scales the noise, and that is why it is recommended to use brighter REFERENCE(s) whenever possible.

If one thinks in terms of spectral energy distribution (SED) for both objects binned in spectral channels, the ideal photometric scaling factors can vary significantly from one spectral channel to the next. One can measure it empirically on the data but only in the spectral bandwidth of the filter. If it is a broadband filter, only an average scaling factor will be applied to the whole polychromatic image which can be thought of a superimposition of many PSFs at different wavelengths. The spectral mismatch between SCIENCE and REFERENCE will thus not only generate extra noise but allow possible under- and over-subtraction at various spatial locations of the PSF. Over-subtraction leads to negative fluxes and affects the estimation of the contrast and hence the detection limits. If one of the objects has strong emission features in its spectrum in the spectral bandwidth that is considered, the effect can be dramatic.

The Exposure Time Calculator (ETC) calculates the flux for each object through a given filter, accounting for the spectral type (or user-provided spectrum). However, the ETC considers the PSF profiles to be exactly the same and hence does not account for the loss in sensitivity due to under- and over-subtraction caused by a spectral mismatch.

This effect is assumed to be negligible above ~5 µm (hence for MIRI). Also, the effect will be obviously stronger closer to the center of the PSF and/or where the coronagraphic 3-D profile has structures (i.e., <10 λ/D). Further out, in the background-limited regime, the effect will be minor. At longer wavelengths, the background-limited regime takes over quickly from the speckle-limited regime where the effect can be substantial.

Effect for NIRCam coronagraphy

The NIRCam team has evaluated the effect of spectral mismatch on sensitivity for separations between 0.5” and 2” from a central object.  These calculations were performed using pyNRC, a Python-based tool making use of WebbPSF. Figures 1–3 show the results for 3 of the most common filters (F200W, F322W2, and F444W) for NIRCam coronagraphic imaging with round occulting masks.

Note: these calculations only account for the effect of spectral mismatch between a science target (vertical axis) and a PSF reference (horizontal axis). They suppose that everything else is optimal (i.e., no thermal drift-induced wavefront errors, no misregistration). Therefore this loss of sensitivity should be thought of as the "best case scenario" if everything else is well mitigated thanks to good observing and PSF subtraction strategies. It is probably safe to assume these results are reliable beyond 1" separation, as inside this region other effects will dominate any spectral mismatch effects. Nevertheless, in many cases the loss of sensitivity due to spectral mismatch may be acceptable and constraints on the spectral type may be relaxed in favor of suitable reference stars that are brighter and/or closer on the sky.

Web users may click on the figures for a larger view.

Figure 1. Estimated sensitivity losses due to spectral mismatch for NIRCam coronagraphy in F200W

Click on the figure for a larger view.

Estimated average sensitivity losses (in magnitudes) due to spectral mismatch between a science target (vertical axis) and a PSF reference (horizontal axis) for separations 0.5”–2” with NIRCam coronagraphy over the whole spectral bandwidth of the F200W filter and using the round occulting mask MASK210R. Other sources of sensitivity loss (including thermal drift and mis-registration) are not considered here.

Figure 2. Estimated sensitivity losses due to spectral mismatch for NIRCam coronagraphy in F322W2

Click on the figure for a larger view.

Estimated average sensitivity losses as in Figure 1, but for F322W2 and the round occulting mask MASK335R.

Figure 3. Estimated sensitivity losses due to spectral mismatch for NIRCam coronagraphy in F444W

Click on the figure for a larger view.

Estimated average sensitivity losses as in Figure 1, but for F444W and the round occulting mask MASK430R.

Selecting PSF reference stars with Simbad

Words in bold are GUI menus/
panels or data software packages; 
bold italics are buttons in GUI
tools or package parameters.

Using Simbad's Query by coordinates form, users can enter the coordinates of the science target and look exhaustively, to the catalogs' sensitivity limit, in a region surrounding the target of interest. From the returned table, one can sort by distance (in arcseconds), spectral type, magnitude and eventually narrow down the search, iteratively, to find the best suited PSF reference stars. Since JWST is mainly an infrared telescope, users should enable the search to include J, H, and K band magnitudes (that are most convenient for comparison with the science target); this is done prior to executing the query by clicking on the Output options button that opens the Options and output parameters form where these magnitudes can be selected at the Fluxes/Magnitudes parameter.

Using Simbad's Query by criteria form, users can search specific ranges of right ascension, declination, magnitude, and even spectral type. Here is an example query to search for a PSF reference star in the vicinity of Beta Pictoris with similar properties: 

rah >= 04 & rah <= 07 & dec > -70 & dec < -40 & Kmag <= 4 & sptypes <= 'A9'

The returned results are:

Number of objects : 6

#|       identifier        |typ|  coord1 (ICRS,J2000/2000)   |Mag V |Mag K |  spec. type   |#bib|#not
1|* bet Pic                |*  |05 47 17.08769 -51 03 59.4412| 3.86 | 3.48 |A6V            |1149|   1
2|* del Dor                |*  |05 44 46.37811 -65 44 07.9011| 4.36 | 3.71 |A7V            |  50|   0
3|* alf Dor                |a2*|04 33 59.77719 -55 02 41.9243| 3.28 | 3.52 |B8IIIpSi       |  46|   0
4|* nu. Pup                |Ce*|06 37 45.67135 -43 11 45.3602| 3.17 | 3.39 |B8III          |  35|   0
5|* alf Car                |*  |06 23 57.10988 -52 41 44.3810|-0.74 |-1.35 |A9II           | 149|   0
6|* alf Pic                |PM*|06 48 11.45512 -61 56 29.0008| 3.30 | 2.570|A8VnkA6        |  67|   0


  • In order to get such output with Simbad, one needs to select the Return/display option. 
  • Mag K (as well as many other parameters) can be selected using the Output options page of Simbad, available from the top menu.
  • The Python package Astroquery should allow users to perform similar Simbad and/or VizieR queries in a command line or batch manner.

Selecting PSF reference stars with SearchCal 

The Jean-Marie Mariotti Center (JMMC) has created tools for the community including SearchCal, a GUI that allows users to select suitable, non-resolved calibrator targets matching various criteria. While SearchCal was designed for long-baseline optical/IR interferometry (hence the squared visibility criterion), it can easily be used to match JWST HCI needs. It offers a practical and graphical way to narrow down a search of PSF reference stars from a catalog of 2.5 million pre-selected stars (with computed and/or measured stellar diameters).

To use SearchCal for your JWST HCI needs:

  • Query your science target in the Name field of the GUI (top, center)
  • Select K as the Magnitude Band in the Instrumental Configuration box (top, left). The K band magnitude of the object will automatically be fetched from Simbad.
  • Chose your Scenario in the SearchCal Parameters box (top, right):
    • Faint: All 2.5 million stars from the catalog are browsed and returned in a circular patch with a maximum radius of 3,600 arcmin (60º).
    • Bright: The research field is then a rectangular box, with a maximum size of 240.0 min (60º) in right ascension and 30.0º in declination, The stars are on average brighter and have a known spectral type. Since we care about the spectral type for the reasons explained in the section above, the Bright scenario is preferred as the first iteration, unless no suitable star is found.
  • Click on the Get Calibrators button (top right corner below the SearchCal Parameters box); this will produce a list of objects in the Found Calibrators middle sub-panel
  • Use the Filters (bottom sub-panel) to narrow down the search while focusing on the distance (dist column to the left) and the spectral type (SpType column) in the results listing. The most useful filters to enable for JWST HCI are:
    • Reject Invalid Object
    • Reject Multiplicity though in many cases, the possible additional component(s) will be too far away (>2") to affect the PSF subtraction
    • Reject Spectral Types (too different from your science target)
    • It can also be a good idea to select stars which are as bright as or brighter than your science target.

In some cases (away from the Galactic plane), the search will not return many stars. In that instance, you may want to relax some criteria and/or cross check with Simbad.

Figure 4. SearchCal query for a suitable PSF reference star for Beta Pictoris

Click on the figure for a larger view.

Screen capture of a SearchCal query to look for a suitable PSF reference star for Beta Pictoris. In this case, Alpha Pictoris has previously been used as a PSF calibrator for many HST and ground-based programs to subtract from Beta Pictoris. SearchCal shows that Alpha Pic is indeed the closest (~13º apart) A-type star that is brighter than Beta Pic.


Lafrenière, D., et al., 
2007, ApJ, 660, 770
A New Algorithm for Point-spread Function Subtraction in High-Contrast Imaging: A demonstration with Angular Differential Imaging

LaJoie, C-P, et al. 2016, SPIE 9904 Space Telescopes and Instrumentation: Optical, Infrared, and Millimeter Wave
Small-grid dithers for the JWST coronagraphs

Soummer, R., Pueyo, L. Larkin, J.,  2012,  ApJL, 755, L28
Detection and Characterization of 
Exoplanets and Disks Using Projections on Karhunen-Loeve Eigenimages

Soummer, R., et al., 2014, SPIE 9143 JWST-STScI-004142
Small-Grid Dithering Strategy for Improved Coronagraphic Performance with JWST

Leisenring, J. and contributors (University of Arizona (2015-2018)
pyNRC - Python ETC and Simulator for JWST NIRCam

Duvert, G. 2016 
VizieR Online Data Catalog: JMDC : JMMC Measured Stellar Diameters Catalogue

Chelli A., Duvert G., Bourgès L. et al., 2016, A&A, 589, 112
Pseudomagnitudes and differential surface brightness: Application to the apparent diameter of stars

Notable updates
Originally published